EDB Postgres Advanced Server Guide

EDB Postgres Advanced Server 10

November 6, 2017
 EDB Postgres Advanced Server Guide
by EnterpriseDB Corporation
Copyright 2014 - 2017 EnterpriseDB Corporation

EnterpriseDB Corporation, 34 Crosby Drive, Suite 100, Bedford, MA 01730, USA

T +1 781 357 3390 F +1 978 589 5701 E info@enterprisedb.com www.enterprisedb.com

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 2

 EDB Postgres Advanced Server Guide

Table of Contents
1 Introduction ................................................................................................................. 8
1.1 Whats New ...................................................................................................... 10
1.2 Typographical Conventions Used in this Guide ............................................... 11
1.3 Other Conventions Used in this Guide ............................................................. 12
1.4 About the Examples Used in this Guide ........................................................... 13
1.4.1 Sample Database Description ....................................................................... 13
2 Enhanced Compatibility Features ............................................................................. 22
2.1 Enabling Compatibility Features ...................................................................... 23
2.2 Stored Procedural Language ............................................................................. 23
2.3 Optimizer Hints................................................................................................. 24
2.4 Data Dictionary Views ...................................................................................... 24
2.5 dblink_ora ......................................................................................................... 25
2.6 Profile Management .......................................................................................... 25
2.7 Built-In Packages .............................................................................................. 26
2.8 Open Client Library .......................................................................................... 28
2.9 Utilities.............................................................................................................. 29
2.10 ECPGPlus ......................................................................................................... 31
2.11 Table Partitioning.............................................................................................. 32
3 Database Administration ........................................................................................... 33
3.1 Configuration Parameters ................................................................................. 33
3.1.1 Setting Configuration Parameters ................................................................. 34
3.1.2 Summary of Configuration Parameters ........................................................ 36
3.1.3 Configuration Parameters by Functionality .................................................. 54
3.1.3.1 Top Performance Related Parameters ................................................... 55
3.1.3.2 Resource Usage / Memory .................................................................... 66
3.1.3.3 Resource Usage / EDB Resource Manager .......................................... 71
3.1.3.4 Query Tuning ........................................................................................ 73
3.1.3.5 Query Tuning / Planner Method Configuration .................................... 73
3.1.3.6 Reporting and Logging / What to Log .................................................. 76
3.1.3.7 Auditing Settings .................................................................................. 78
3.1.3.8 Client Connection Defaults / Locale and Formatting ........................... 83
3.1.3.9 Client Connection Defaults / Statement Behavior ................................ 83
3.1.3.10 Client Connection Defaults / Other Defaults .................................... 84
3.1.3.11 Compatibility Options ....................................................................... 86
3.1.3.12 Customized Options .......................................................................... 94
3.1.3.13 Ungrouped....................................................................................... 103
3.2 Index Advisor.................................................................................................. 106
3.2.1 Index Advisor Components ........................................................................ 107
3.2.2 Index Advisor Configuration ...................................................................... 108
3.2.3 Using Index Advisor ................................................................................... 111
3.2.3.1 Using the pg_advise_index Utility...................................................... 111
3.2.3.2 Using Index Advisor at the psql Command Line................................ 113
3.2.4 Reviewing the Index Advisor Recommendations....................................... 115
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 3
 EDB Postgres Advanced Server Guide

3.2.4.1 Using the show_index_recommendations() Function ........................ 115

3.2.4.2 Querying the index_advisor_log Table ............................................... 116
3.2.4.3 Querying the index_recommendations View...................................... 118
3.2.5 Limitations .................................................................................................. 119
3.3 SQL Profiler .................................................................................................... 121
3.4 pgsnmpd .......................................................................................................... 123
3.4.1 Configuring pgsnmpd ................................................................................. 123
3.4.2 Setting the Listener Address ....................................................................... 124
3.4.3 Invoking pgsnmpd ...................................................................................... 124
3.4.4 Viewing pgsnmpd Help .............................................................................. 125
3.4.5 Requesting Information from pgsnmpd ...................................................... 125
3.5 EDB Audit Logging ........................................................................................ 126
3.5.1 Audit Logging Configuration Parameters ................................................... 126
3.5.2 Selecting SQL Statements to Audit ............................................................ 128
3.5.2.1 Data Definition Language and Data Control Language Statements ... 129
3.5.2.2 Data Manipulation Language Statements ........................................... 134
3.5.3 Enabling Audit Logging ............................................................................. 137
3.5.4 Audit Log File ............................................................................................. 141
3.5.5 Filtering Audit Logs Using Error Codes ..................................................... 146
3.5.6 Filtering Audit Logs Using Command Tags ............................................... 146
3.6 Unicode Collation Algorithm ......................................................................... 148
3.6.1 Basic Unicode Collation Algorithm Concepts ............................................ 149
3.6.2 International Components for Unicode ....................................................... 150
3.6.2.1 Locale Collations ................................................................................ 150
3.6.2.2 Collation Attributes............................................................................. 151
3.6.3 Creating an ICU Collation .......................................................................... 154
3.6.3.1 CREATE COLLATION ..................................................................... 154
3.6.3.2 CREATE DATABASE ....................................................................... 156
3.6.3.3 initdb ................................................................................................... 157
3.6.4 Using a Collation ........................................................................................ 159
3.7 Customizable WAL Segment File Size .......................................................... 163
4 Security ................................................................................................................... 165
4.1 Protecting Against SQL Injection Attacks ...................................................... 165
4.1.1 SQL/Protect Overview ................................................................................ 166
4.1.1.1 Types of SQL Injection Attacks ......................................................... 166
4.1.1.2 Monitoring SQL Injection Attacks ..................................................... 167
4.1.2 Configuring SQL/Protect ............................................................................ 170
4.1.2.1 Selecting Roles to Protect ................................................................... 172
4.1.2.2 Monitoring Protected Roles ................................................................ 174
4.1.3 Common Maintenance Operations ............................................................. 181
4.1.3.1 Adding a Role to the Protected Roles List .......................................... 181
4.1.3.2 Removing a Role From the Protected Roles List................................ 181
4.1.3.3 Setting the Types of Protection for a Role .......................................... 182
4.1.3.4 Removing a Relation From the Protected Relations List.................... 183
4.1.3.5 Deleting Statistics ............................................................................... 183
4.1.3.6 Deleting Offending Queries ................................................................ 184

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 4

 EDB Postgres Advanced Server Guide

4.1.3.7 Disabling and Enabling Monitoring.................................................... 185

4.1.4 Backing Up and Restoring a SQL/Protect Database................................... 187
4.1.4.1 Object Identification Numbers in SQL/Protect Tables ....................... 187
4.1.4.2 Backing Up the Database .................................................................... 188
4.1.4.3 Restoring From the Backup Files ....................................................... 188
4.2 Virtual Private Database ................................................................................. 193
4.3 sslutils ............................................................................................................. 194
4.3.1 openssl_rsa_generate_key........................................................................... 194
4.3.2 openssl_rsa_key_to_csr .............................................................................. 194
4.3.3 openssl_csr_to_crt....................................................................................... 195
4.3.4 openssl_rsa_generate_crl ............................................................................ 196
5 EDB Resource Manager ......................................................................................... 197
5.1 Creating and Managing Resource Groups ...................................................... 198
5.1.1 CREATE RESOURCE GROUP................................................................. 198
5.1.2 ALTER RESOURCE GROUP ................................................................... 199
5.1.3 DROP RESOURCE GROUP ..................................................................... 201
5.1.4 Assigning a Process to a Resource Group .................................................. 201
5.1.5 Removing a Process from a Resource Group ............................................. 202
5.1.6 Monitoring Processes in Resource Groups ................................................. 203
5.2 CPU Usage Throttling..................................................................................... 205
5.2.1 Setting the CPU Rate Limit for a Resource Group ..................................... 205
5.2.2 Example Single Process in a Single Group.............................................. 206
5.2.3 Example Multiple Processes in a Single Group ....................................... 207
5.2.4 Example Multiple Processes in Multiple Groups .................................... 209
5.3 Dirty Buffer Throttling ................................................................................... 212
5.3.1 Setting the Dirty Rate Limit for a Resource Group .................................... 212
5.3.2 Example Single Process in a Single Group.............................................. 214
5.3.3 Example Multiple Processes in a Single Group ....................................... 215
5.3.4 Example Multiple Processes in Multiple Groups .................................... 217
5.4 System Catalogs .............................................................................................. 221
5.4.1 edb_all_resource_groups ............................................................................ 221
5.4.2 edb_resource_group .................................................................................... 221
6 libpq C Library........................................................................................................ 222
6.1.1 Using libpq with EnterpriseDB SPL ........................................................... 222
6.1.2 REFCURSOR Support................................................................................ 222
6.1.3 Array Binding ............................................................................................. 229
6.1.3.1 PQBulkStart ........................................................................................ 229
6.1.3.2 PQexecBulk ........................................................................................ 229
6.1.3.3 PQBulkFinish ...................................................................................... 230
6.1.3.4 PQexecBulkPrepared .......................................................................... 230
6.1.3.5 Example Code (Using PQBulkStart, PQexecBulk, PQBulkFinish) ... 231
6.1.3.6 Example Code (Using PQexecBulkPrepared) .................................... 232
7 Debugger ................................................................................................................. 233
7.1.1 Configuring the Debugger .......................................................................... 233
7.1.2 Starting the Debugger ................................................................................. 234
7.1.3 The View Data Options Window................................................................ 236

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 5

 EDB Postgres Advanced Server Guide

7.1.4 Main Debugger Window............................................................................. 238

7.1.4.1 The Program Body Pane ..................................................................... 239
7.1.4.2 The Stack Pane .................................................................................... 240
7.1.4.3 The Output Pane ................................................................................. 243
7.1.4.4 The Status Bar ..................................................................................... 244
7.1.5 Debugging a Program ................................................................................. 245
7.1.5.1 Stepping Through the Code ................................................................ 245
7.1.5.2 Using Breakpoints ............................................................................... 246
7.1.5.3 Setting a Global Breakpoint for In-Context Debugging ..................... 249
7.1.5.4 Exiting the Debugger .......................................................................... 255
8 Performance Analysis and Tuning .......................................................................... 256
8.1 Dynatune ......................................................................................................... 256
8.1.1 edb_dynatune .............................................................................................. 256
8.1.2 edb_dynatune_profile ................................................................................. 257
8.2 Infinite Cache .................................................................................................. 258
8.2.1 Installing Infinite Cache .............................................................................. 262
8.2.2 Configuring the Infinite Cache Server ........................................................ 264
8.2.2.1 Modifying Infinite Cache Settings ...................................................... 264
8.2.2.2 Enabling Infinite Cache ...................................................................... 264
8.2.2.3 Controlling the Infinite Cache Server ................................................. 267
8.2.3 Dynamically Modifying Infinite Cache Server Nodes ............................... 268
8.2.4 Controlling the edb-icache Daemons .......................................................... 269
8.2.4.1 Command Line Options ...................................................................... 269
8.2.4.2 edb-icache-tool .................................................................................... 271
8.2.5 Warming the edb-icache Servers ................................................................ 273
8.2.5.1 The edb_icache_warm() Function ...................................................... 273
8.2.5.2 Using the edb_icache_warm Utility.................................................... 274
8.2.6 Retrieving Statistics from Infinite Cache .................................................... 275
8.2.6.1 Using edb_icache_stats() .................................................................... 275
8.2.6.2 edb_icache_server_list ........................................................................ 276
8.2.7 Retrieving Table Statistics .......................................................................... 278
8.2.7.1 pg_statio_all_tables............................................................................. 278
8.2.7.2 pg_statio_sys_tables ........................................................................... 280
8.2.7.3 pg_statio_user_tables .......................................................................... 280
8.2.7.4 pg_statio_all_indexes .......................................................................... 280
8.2.7.5 pg_statio_sys_indexes......................................................................... 282
8.2.7.6 pg_statio_user_indexes ....................................................................... 282
8.2.8 edb_icache_server_enable() ........................................................................ 283
8.2.9 Infinite Cache Log Entries .......................................................................... 284
8.2.10 Allocating Memory to the Cache Servers ............................................... 285
9 EDB Clone Schema ................................................................................................ 286
9.1 Setup Process .................................................................................................. 288
9.1.1 Installing Extensions and PL/Perl ............................................................... 288
9.1.2 Setting Configuration Parameters ............................................................... 290
9.1.2.1 Performance Configuration Parameters .............................................. 290
9.1.2.2 Status Logging .................................................................................... 291

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 6

 EDB Postgres Advanced Server Guide

9.1.3 Installing EDB Clone Schema .................................................................... 292

9.1.4 Creating the Foreign Servers and User Mappings ...................................... 292
9.1.4.1 Foreign Server and User Mapping for Local Cloning Functions ....... 293
9.1.4.2 Foreign Server and User Mapping for Remote Cloning Functions .... 294
9.2 EDB Clone Schema Functions........................................................................ 297
9.2.1 localcopyschema ......................................................................................... 297
9.2.2 localcopyschema_nb ................................................................................... 301
9.2.3 remotecopyschema ...................................................................................... 304
9.2.4 remotecopyschema_nb ................................................................................ 308
9.2.5 process_status_from_log............................................................................. 311
9.2.6 remove_log_file_and_job ........................................................................... 311
10 PL/Java.................................................................................................................... 313
10.1 Installing PL/Java on Linux ............................................................................ 313
10.2 Installing PL/Java on Windows ...................................................................... 314
10.3 Using PL/Java ................................................................................................. 315
11 Enhanced SQL Features .......................................................................................... 318
11.1 COMMENT .................................................................................................... 318
12 System Catalog Tables ............................................................................................ 323
12.1 edb_dir ............................................................................................................ 323
12.2 edb_all_resource_groups ................................................................................ 323
12.3 edb_password_history..................................................................................... 324
12.4 edb_policy ....................................................................................................... 324
12.5 edb_profile ...................................................................................................... 325
12.6 edb_resource_group ........................................................................................ 326
12.7 edb_variable .................................................................................................... 326
12.8 pg_synonym .................................................................................................... 327
12.9 product_component_version ........................................................................... 327
13 Advanced Server Keywords ................................................................................... 328

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 7

 EDB Postgres Advanced Server Guide

1 Introduction
This guide describes the features of EDB Postgres Advanced Server (Advanced Server).

Advanced Server adds extended functionality to the open-source PostgreSQL database.

The extended functionality supports database administration, enhanced SQL capabilities,
database and application security, performance monitoring and analysis, and application
development utilities. This guide documents those features that are exclusive to
Advanced Server:

Enhanced Compatibility Features. Chapter 2 provides an overview of

compatibility features supported by Advanced Server.

Database Administration. Chapter 3 contains information about features and

tools that are helpful to the database administrator.

Index Advisor described in Section 3.2 helps to determine the additional indexes
needed on tables to improve application performance.

SQL Profiler described in Section 3.3 locates and diagnoses poorly running SQL
queries in applications.

pgsnmpd described in Section 3.4 is an SNMP agent that returns hierarchical

monitoring information regarding the current state of Advanced Server.

Security. Chapter 4 contains information about security features supported by

Advanced Server.

SQL/Protect described in Section 4.1 provides protection against SQL injection

attacks.

Virtual Private Database described in Section 4.2 provides fine-grained, row

level access.

sslutils described in Section 4.3 provides SSL certificate generation functions.

EDB Resource Manager. Chapter 5 contains information about the EDB

Resource Manager feature, which provides the capability to control system
resource usage by Advanced Server processes.

Resource Groups described in Section 5.1 shows how to create and maintain the
groups on which resource limits can be defined and to which Advanced Server
processes can be assigned.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 8

 EDB Postgres Advanced Server Guide

CPU Usage Throttling described in Section 5.2 provides a method to control CPU
usage by Advanced Server processes.

Dirty Buffer Throttling described in Section 5.3 provides a method to control the
dirty rate of shared buffers by Advanced Server processes.

The libpq C Library. The libpq C library described in Chapter 6 is the C

application programming interface (API) language for Advanced Server.

The PL Debugger. The PL Debugger described in Chapter 7 is a graphically

oriented debugging tool for PL/pgSQL.

Performance Analysis and Tuning. Chapter 8 contains the various tools for
analyzing and improving application and database server performance.

Dynatune described in Section 8.1 provides a quick and easy means for
configuring Advanced Server depending upon the type of application usage.

Infinite Cache described in Section 8.2 provides for performance improvement

using memory caching. Note: Infinite Cache has been deprecated and may be
removed in a future release. Please contact your EnterpriseDB Account Manager
or mailto:sales@enterprisedb.com for more information.

EDB Clone Schema. Chapter 9 contains information about the EDB Clone
Schema feature, which provides the capability to copy a schema and its database
objects within a single database or from one database to another database.

PL/Java. PL/Java described in Chapter 10 is a package providing access to Java

stored functions, procedures, and triggers with the JDBC interface.

Enhanced SQL Features. Chapter 11 contains information on enhanced SQL

functionality that provides additional flexibility and convenience.

System Catalog Tables. Chapter 12 contains additional system catalog tables

added for Advanced Server specific database objects.

Advanced Server Keywords. Chapter 13 contains information about the words

that Advanced Server recognizes as keywords.

For information about the features that are shared by Advanced Server and PostgreSQL,
see the PostgreSQL core documentation, available at:

https://www.postgresql.org/docs/10/static/index.html

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 9

 EDB Postgres Advanced Server Guide

1.1 Whats New

The following features have been added to EDB Postgres Advanced Server 9.6 to create
Advanced Server 10:

Advanced Server now includes the following enhancements for EDB Audit
Logging: 1) Separate auditing for each DDL type allows the specification of
individual DDL commands (CREATE, ALTER, or DROP) with or without their
object types (for example, TABLE, VIEW, SEQUENCE, etc.) for which audit logging
is to occur. This refines the types of SQL statements recorded in the audit log file
(see Section 3.5.2.1). 2) Separate auditing for each DML type allows the
specification of individual DML commands (INSERT, UPDATE, DELETE, or
TRUNCATE) for which audit logging is to occur. This provides the same benefit as
separate DDL auditing (see Section 3.5.2.2). 3) Statements to be audited can now
be determined by: a) the database in which auditing is to occur, b) the role
running the session, or c) the combination of the role and the database. This is
accomplished by setting the edb_audit_statement configuration parameter
with the ALTER DATABASE, ALTER ROLE, or ALTER ROLE IN DATABASE
command (see Section 3.5.3). 4) Command tags identify the SQL command that
was executed, and thus aids in scanning the audit log to find entries related to
certain SQL commands (see Section 3.5.6). 5) Addition of the
edb_audit_destination parameter specifies whether audit logging is to be
recorded in the default location under the database cluster or to be handled by the
syslog process (see Section 3.5.1).

Advanced Server now includes the edb_log_every_bulk_value

configuration parameter, which affects both Advanced Server logging and EDB
Audit Logging. Logging a single message for the entire bulk process is now the
default option as opposed to logging every statement involved in bulk processing.
For more information see Section 3.1.3.6.2 or Section 3.5.1.

Advanced Server now includes the EDB Clone Schema extension, which allows
you to make a copy of a schema with its database objects from a given database,
and to insert it into another database, or into the same database (with a different
schema name). The cloning functionality can be run in both an online mode or a
non-blocking, background job mode. Using a background job frees up access to
your terminal while the cloning operation is in progress. Multiple background
worker processes can be used for the cloning process to shorten the amount of
time to complete the operation. For more information, see Chapter 9.

Advanced Server now provides the pg_prewarm module, which implements the
autoprewarm background worker. The autoprewarm worker process
automatically dumps shared buffers to disk before a shutdown. It then prewarms
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 10
 EDB Postgres Advanced Server Guide

the shared buffers the next time the server is started, meaning it loads blocks from
the disk back into the buffer pool. This shortens the warm up times after the
server has been restarted. For more information, see Section 3.1.3.1.16.

Advanced Server now provides the --wal-segsize option for the initdb
utility program. This provides the capability to specify the WAL segment file size
when creating a database cluster instead of using the default size of 16 MB. For
more information, see Section 3.7.

1.2 Typographical Conventions Used in this Guide

Certain typographical conventions are used in this manual to clarify the meaning and
usage of various commands, statements, programs, examples, etc. This section provides a
summary of these conventions.

In the following descriptions a term refers to any word or group of words that may be
language keywords, user-supplied values, literals, etc. A terms exact meaning depends
upon the context in which it is used.

Italic font introduces a new term, typically, in the sentence that defines it for the
first time.
Fixed-width (mono-spaced) font is used for terms that must be given
literally such as SQL commands, specific table and column names used in the
examples, programming language keywords, directory paths and file names,
parameter values, etc. For example postgresql.conf, SELECT * FROM emp;
Italic fixed-width font is used for terms for which the user must
substitute values in actual usage. For example, DELETE FROM table_name;
A vertical pipe | denotes a choice between the terms on either side of the pipe. A
vertical pipe is used to separate two or more alternative terms within square
brackets (optional choices) or braces (one mandatory choice).
Square brackets [ ] denote that one or none of the enclosed term(s) may be
substituted. For example, [ a | b ], means choose one of a or b or neither
of the two.
Braces {} denote that exactly one of the enclosed alternatives must be specified.
For example, { a | b }, means exactly one of a or b must be specified.
Ellipses ... denote that the proceeding term may be repeated. For example, [ a |
b ] ... means that you may have the sequence, b a a b a.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 11

 EDB Postgres Advanced Server Guide

1.3 Other Conventions Used in this Guide

The following is a list of other conventions used throughout this document.

This guide applies to both Linux and Windows systems. Directory paths are
presented in the Linux format with forward slashes. When working on Windows
systems, start the directory path with the drive letter followed by a colon and
substitute back slashes for forward slashes.
Some of the information in this document may apply interchangeably to the
PostgreSQL and EDB Postgres Advanced Server database systems. The term
Advanced Server is used to refer to EDB Postgres Advanced Server. The term
Postgres is used to generically refer to both PostgreSQL and Advanced Server.
When a distinction needs to be made between these two database systems, the
specific names, PostgreSQL or Advanced Server are used.
The installation directory path of the PostgreSQL or Advanced Server products is
referred to as POSTGRES_INSTALL_HOME. For PostgreSQL Linux installations,
this defaults to /opt/PostgreSQL/x.x. For PostgreSQL Windows
installations, this defaults to C:\Program Files\PostgreSQL\x.x. For
Advanced Server Linux installations accomplished using the interactive installer,
this defaults to /opt/edb/asx.x. For Advanced Server Linux installations
accomplished using an RPM package, this defaults to /usr/edb/asx.x. For
Advanced Server Windows installations, this defaults to C:\Program
Files\edb\asx.x. The product version number is represented by x.x.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 12

 EDB Postgres Advanced Server Guide

1.4 About the Examples Used in this Guide

The examples in this guide are shown in the type and background illustrated below.

Examples and output from examples are shown in fixed-width, blue font on a
light blue background.

The examples use the sample tables, dept, emp, and jobhist, created and loaded when
Advanced Server is installed.

The tables and programs in the sample database can be re-created at any time by
executing the following script:

/opt/edb/as10/installer/server/pg-sample.sql.

In addition there is a script in the same directory containing the database objects created
using syntax compatible with Oracle databases. This script file is edb-sample.sql.

The script:

Creates the sample tables and programs in the currently connected database.
Grants all permissions on the tables to the PUBLIC group.

The tables and programs will be created in the first schema of the search path in which
the current user has permission to create tables and procedures. You can display the
search path by issuing the command:

SHOW SEARCH_PATH;

You can use PSQL commands to modify the search path.

1.4.1 Sample Database Description

The sample database represents employees in an organization. It contains three types of

records: employees, departments, and historical records of employees.

Each employee has an identification number, name, hire date, salary, and manager. Some
employees earn a commission in addition to their salary. All employee-related
information is stored in the emp table.

The sample company is regionally diverse, so it tracks the locations of its departments.
Each company employee is assigned to a department. Each department is identified by a
unique department number and a short name. Each department is associated with one
location. All department-related information is stored in the dept table.
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 13
 EDB Postgres Advanced Server Guide

The company also tracks information about jobs held by the employees. Some employees
have been with the company for a long time and have held different positions, received
raises, switched departments, etc. When a change in employee status occurs, the company
records the end date of the former position. A new job record is added with the start date
and the new job title, department, salary, and the reason for the status change. All
employee history is maintained in the jobhist table.

The following is the pg-sample.sql script:

SET datestyle TO 'iso, dmy';

--
-- Script that creates the 'sample' tables, views
-- functions, triggers, etc.
--
-- Start new transaction - commit all or nothing
--
BEGIN;
--
-- Create and load tables used in the documentation examples.
--
-- Create the 'dept' table
--
CREATE TABLE dept (
deptno NUMERIC(2) NOT NULL CONSTRAINT dept_pk PRIMARY KEY,
dname VARCHAR(14) CONSTRAINT dept_dname_uq UNIQUE,
loc VARCHAR(13)
);
--
-- Create the 'emp' table
--
CREATE TABLE emp (
empno NUMERIC(4) NOT NULL CONSTRAINT emp_pk PRIMARY KEY,
ename VARCHAR(10),
job VARCHAR(9),
mgr NUMERIC(4),
hiredate DATE,
sal NUMERIC(7,2) CONSTRAINT emp_sal_ck CHECK (sal > 0),
comm NUMERIC(7,2),
deptno NUMERIC(2) CONSTRAINT emp_ref_dept_fk
REFERENCES dept(deptno)
);
--
-- Create the 'jobhist' table
--
CREATE TABLE jobhist (
empno NUMERIC(4) NOT NULL,
startdate TIMESTAMP(0) NOT NULL,
enddate TIMESTAMP(0),
job VARCHAR(9),
sal NUMERIC(7,2),
comm NUMERIC(7,2),
deptno NUMERIC(2),
chgdesc VARCHAR(80),
CONSTRAINT jobhist_pk PRIMARY KEY (empno, startdate),
CONSTRAINT jobhist_ref_emp_fk FOREIGN KEY (empno)
REFERENCES emp(empno) ON DELETE CASCADE,
CONSTRAINT jobhist_ref_dept_fk FOREIGN KEY (deptno)
REFERENCES dept (deptno) ON DELETE SET NULL,

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 14

 EDB Postgres Advanced Server Guide

CONSTRAINT jobhist_date_chk CHECK (startdate <= enddate)

);
--
-- Create the 'salesemp' view
--
CREATE OR REPLACE VIEW salesemp AS
SELECT empno, ename, hiredate, sal, comm FROM emp WHERE job = 'SALESMAN';
--
-- Sequence to generate values for function 'new_empno'.
--
CREATE SEQUENCE next_empno START WITH 8000 INCREMENT BY 1;
--
-- Issue PUBLIC grants
--
--GRANT ALL ON emp TO PUBLIC;
--GRANT ALL ON dept TO PUBLIC;
--GRANT ALL ON jobhist TO PUBLIC;
--GRANT ALL ON salesemp TO PUBLIC;
--GRANT ALL ON next_empno TO PUBLIC;
--
-- Load the 'dept' table
--
INSERT INTO dept VALUES (10,'ACCOUNTING','NEW YORK');
INSERT INTO dept VALUES (20,'RESEARCH','DALLAS');
INSERT INTO dept VALUES (30,'SALES','CHICAGO');
INSERT INTO dept VALUES (40,'OPERATIONS','BOSTON');
--
-- Load the 'emp' table
--
INSERT INTO emp VALUES (7369,'SMITH','CLERK',7902,'17-DEC-80',800,NULL,20);
INSERT INTO emp VALUES (7499,'ALLEN','SALESMAN',7698,'20-FEB-
81',1600,300,30);
INSERT INTO emp VALUES (7521,'WARD','SALESMAN',7698,'22-FEB-81',1250,500,30);
INSERT INTO emp VALUES (7566,'JONES','MANAGER',7839,'02-APR-
81',2975,NULL,20);
INSERT INTO emp VALUES (7654,'MARTIN','SALESMAN',7698,'28-SEP-
81',1250,1400,30);
INSERT INTO emp VALUES (7698,'BLAKE','MANAGER',7839,'01-MAY-
81',2850,NULL,30);
INSERT INTO emp VALUES (7782,'CLARK','MANAGER',7839,'09-JUN-
81',2450,NULL,10);
INSERT INTO emp VALUES (7788,'SCOTT','ANALYST',7566,'19-APR-
87',3000,NULL,20);
INSERT INTO emp VALUES (7839,'KING','PRESIDENT',NULL,'17-NOV-
81',5000,NULL,10);
INSERT INTO emp VALUES (7844,'TURNER','SALESMAN',7698,'08-SEP-81',1500,0,30);
INSERT INTO emp VALUES (7876,'ADAMS','CLERK',7788,'23-MAY-87',1100,NULL,20);
INSERT INTO emp VALUES (7900,'JAMES','CLERK',7698,'03-DEC-81',950,NULL,30);
INSERT INTO emp VALUES (7902,'FORD','ANALYST',7566,'03-DEC-81',3000,NULL,20);
INSERT INTO emp VALUES (7934,'MILLER','CLERK',7782,'23-JAN-82',1300,NULL,10);
--
-- Load the 'jobhist' table
--
INSERT INTO jobhist VALUES (7369,'17-DEC-80',NULL,'CLERK',800,NULL,20,'New
Hire');
INSERT INTO jobhist VALUES (7499,'20-FEB-81',NULL,'SALESMAN',1600,300,30,'New
Hire');
INSERT INTO jobhist VALUES (7521,'22-FEB-81',NULL,'SALESMAN',1250,500,30,'New
Hire');
INSERT INTO jobhist VALUES (7566,'02-APR-81',NULL,'MANAGER',2975,NULL,20,'New
Hire');
INSERT INTO jobhist VALUES (7654,'28-SEP-
81',NULL,'SALESMAN',1250,1400,30,'New Hire');
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 15
 EDB Postgres Advanced Server Guide

INSERT INTO jobhist VALUES (7698,'01-MAY-81',NULL,'MANAGER',2850,NULL,30,'New

Hire');
INSERT INTO jobhist VALUES (7782,'09-JUN-81',NULL,'MANAGER',2450,NULL,10,'New
Hire');
INSERT INTO jobhist VALUES (7788,'19-APR-87','12-APR-
88','CLERK',1000,NULL,20,'New Hire');
INSERT INTO jobhist VALUES (7788,'13-APR-88','04-MAY-
89','CLERK',1040,NULL,20,'Raise');
INSERT INTO jobhist VALUES (7788,'05-MAY-
90',NULL,'ANALYST',3000,NULL,20,'Promoted to Analyst');
INSERT INTO jobhist VALUES (7839,'17-NOV-
81',NULL,'PRESIDENT',5000,NULL,10,'New Hire');
INSERT INTO jobhist VALUES (7844,'08-SEP-81',NULL,'SALESMAN',1500,0,30,'New
Hire');
INSERT INTO jobhist VALUES (7876,'23-MAY-87',NULL,'CLERK',1100,NULL,20,'New
Hire');
INSERT INTO jobhist VALUES (7900,'03-DEC-81','14-JAN-
83','CLERK',950,NULL,10,'New Hire');
INSERT INTO jobhist VALUES (7900,'15-JAN-
83',NULL,'CLERK',950,NULL,30,'Changed to Dept 30');
INSERT INTO jobhist VALUES (7902,'03-DEC-81',NULL,'ANALYST',3000,NULL,20,'New
Hire');
INSERT INTO jobhist VALUES (7934,'23-JAN-82',NULL,'CLERK',1300,NULL,10,'New
Hire');
--
-- Populate statistics table and view (pg_statistic/pg_stats)
--
ANALYZE dept;
ANALYZE emp;
ANALYZE jobhist;
--
-- Function that lists all employees' numbers and names
-- from the 'emp' table using a cursor.
--
CREATE OR REPLACE FUNCTION list_emp() RETURNS VOID
AS $$
DECLARE
v_empno NUMERIC(4);
v_ename VARCHAR(10);
emp_cur CURSOR FOR
SELECT empno, ename FROM emp ORDER BY empno;
BEGIN
OPEN emp_cur;
RAISE INFO 'EMPNO ENAME';
RAISE INFO '----- -------';
LOOP
FETCH emp_cur INTO v_empno, v_ename;
EXIT WHEN NOT FOUND;
RAISE INFO '% %', v_empno, v_ename;
END LOOP;
CLOSE emp_cur;
RETURN;
END;
$$ LANGUAGE 'plpgsql';
--
-- Function that selects an employee row given the employee
-- number and displays certain columns.
--
CREATE OR REPLACE FUNCTION select_emp (
p_empno NUMERIC
) RETURNS VOID
AS $$
DECLARE
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 16
 EDB Postgres Advanced Server Guide

v_ename emp.ename%TYPE;
v_hiredate emp.hiredate%TYPE;
v_sal emp.sal%TYPE;
v_comm emp.comm%TYPE;
v_dname dept.dname%TYPE;
v_disp_date VARCHAR(10);
BEGIN
SELECT INTO
v_ename, v_hiredate, v_sal, v_comm, v_dname
ename, hiredate, sal, COALESCE(comm, 0), dname
FROM emp e, dept d
WHERE empno = p_empno
AND e.deptno = d.deptno;
IF NOT FOUND THEN
RAISE INFO 'Employee % not found', p_empno;
RETURN;
END IF;
v_disp_date := TO_CHAR(v_hiredate, 'MM/DD/YYYY');
RAISE INFO 'Number : %', p_empno;
RAISE INFO 'Name : %', v_ename;
RAISE INFO 'Hire Date : %', v_disp_date;
RAISE INFO 'Salary : %', v_sal;
RAISE INFO 'Commission: %', v_comm;
RAISE INFO 'Department: %', v_dname;
RETURN;
EXCEPTION
WHEN OTHERS THEN
RAISE INFO 'The following is SQLERRM : %', SQLERRM;
RAISE INFO 'The following is SQLSTATE: %', SQLSTATE;
RETURN;
END;
$$ LANGUAGE 'plpgsql';
--
-- A RECORD type used to format the return value of
-- function, 'emp_query'.
--
CREATE TYPE emp_query_type AS (
empno NUMERIC,
ename VARCHAR(10),
job VARCHAR(9),
hiredate DATE,
sal NUMERIC
);
--
-- Function that queries the 'emp' table based on
-- department number and employee number or name. Returns
-- employee number and name as INOUT parameters and job,
-- hire date, and salary as OUT parameters. These are
-- returned in the form of a record defined by
-- RECORD type, 'emp_query_type'.
--
CREATE OR REPLACE FUNCTION emp_query (
IN p_deptno NUMERIC,
INOUT p_empno NUMERIC,
INOUT p_ename VARCHAR,
OUT p_job VARCHAR,
OUT p_hiredate DATE,
OUT p_sal NUMERIC
)
AS $$
BEGIN
SELECT INTO
p_empno, p_ename, p_job, p_hiredate, p_sal
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 17
 EDB Postgres Advanced Server Guide

empno, ename, job, hiredate, sal

FROM emp
WHERE deptno = p_deptno
AND (empno = p_empno
OR ename = UPPER(p_ename));
END;
$$ LANGUAGE 'plpgsql';
--
-- Function to call 'emp_query_caller' with IN and INOUT
-- parameters. Displays the results received from INOUT and
-- OUT parameters.
--
CREATE OR REPLACE FUNCTION emp_query_caller() RETURNS VOID
AS $$
DECLARE
v_deptno NUMERIC;
v_empno NUMERIC;
v_ename VARCHAR;
v_rows INTEGER;
r_emp_query EMP_QUERY_TYPE;
BEGIN
v_deptno := 30;
v_empno := 0;
v_ename := 'Martin';
r_emp_query := emp_query(v_deptno, v_empno, v_ename);
RAISE INFO 'Department : %', v_deptno;
RAISE INFO 'Employee No: %', (r_emp_query).empno;
RAISE INFO 'Name : %', (r_emp_query).ename;
RAISE INFO 'Job : %', (r_emp_query).job;
RAISE INFO 'Hire Date : %', (r_emp_query).hiredate;
RAISE INFO 'Salary : %', (r_emp_query).sal;
RETURN;
EXCEPTION
WHEN OTHERS THEN
RAISE INFO 'The following is SQLERRM : %', SQLERRM;
RAISE INFO 'The following is SQLSTATE: %', SQLSTATE;
RETURN;
END;
$$ LANGUAGE 'plpgsql';
--
-- Function to compute yearly compensation based on semimonthly
-- salary.
--
CREATE OR REPLACE FUNCTION emp_comp (
p_sal NUMERIC,
p_comm NUMERIC
) RETURNS NUMERIC
AS $$
BEGIN
RETURN (p_sal + COALESCE(p_comm, 0)) * 24;
END;
$$ LANGUAGE 'plpgsql';
--
-- Function that gets the next number from sequence, 'next_empno',
-- and ensures it is not already in use as an employee number.
--
CREATE OR REPLACE FUNCTION new_empno() RETURNS INTEGER
AS $$
DECLARE
v_cnt INTEGER := 1;
v_new_empno INTEGER;
BEGIN
WHILE v_cnt > 0 LOOP
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 18
 EDB Postgres Advanced Server Guide

SELECT INTO v_new_empno nextval('next_empno');

SELECT INTO v_cnt COUNT(*) FROM emp WHERE empno = v_new_empno;
END LOOP;
RETURN v_new_empno;
END;
$$ LANGUAGE 'plpgsql';
--
-- Function that adds a new clerk to table 'emp'.
--
CREATE OR REPLACE FUNCTION hire_clerk (
p_ename VARCHAR,
p_deptno NUMERIC
) RETURNS NUMERIC
AS $$
DECLARE
v_empno NUMERIC(4);
v_ename VARCHAR(10);
v_job VARCHAR(9);
v_mgr NUMERIC(4);
v_hiredate DATE;
v_sal NUMERIC(7,2);
v_comm NUMERIC(7,2);
v_deptno NUMERIC(2);
BEGIN
v_empno := new_empno();
INSERT INTO emp VALUES (v_empno, p_ename, 'CLERK', 7782,
CURRENT_DATE, 950.00, NULL, p_deptno);
SELECT INTO
v_empno, v_ename, v_job, v_mgr, v_hiredate, v_sal, v_comm, v_deptno
empno, ename, job, mgr, hiredate, sal, comm, deptno
FROM emp WHERE empno = v_empno;
RAISE INFO 'Department : %', v_deptno;
RAISE INFO 'Employee No: %', v_empno;
RAISE INFO 'Name : %', v_ename;
RAISE INFO 'Job : %', v_job;
RAISE INFO 'Manager : %', v_mgr;
RAISE INFO 'Hire Date : %', v_hiredate;
RAISE INFO 'Salary : %', v_sal;
RAISE INFO 'Commission : %', v_comm;
RETURN v_empno;
EXCEPTION
WHEN OTHERS THEN
RAISE INFO 'The following is SQLERRM : %', SQLERRM;
RAISE INFO 'The following is SQLSTATE: %', SQLSTATE;
RETURN -1;
END;
$$ LANGUAGE 'plpgsql';
--
-- Function that adds a new salesman to table 'emp'.
--
CREATE OR REPLACE FUNCTION hire_salesman (
p_ename VARCHAR,
p_sal NUMERIC,
p_comm NUMERIC
) RETURNS NUMERIC
AS $$
DECLARE
v_empno NUMERIC(4);
v_ename VARCHAR(10);
v_job VARCHAR(9);
v_mgr NUMERIC(4);
v_hiredate DATE;
v_sal NUMERIC(7,2);
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 19
 EDB Postgres Advanced Server Guide

v_comm NUMERIC(7,2);
v_deptno NUMERIC(2);
BEGIN
v_empno := new_empno();
INSERT INTO emp VALUES (v_empno, p_ename, 'SALESMAN', 7698,
CURRENT_DATE, p_sal, p_comm, 30);
SELECT INTO
v_empno, v_ename, v_job, v_mgr, v_hiredate, v_sal, v_comm, v_deptno
empno, ename, job, mgr, hiredate, sal, comm, deptno
FROM emp WHERE empno = v_empno;
RAISE INFO 'Department : %', v_deptno;
RAISE INFO 'Employee No: %', v_empno;
RAISE INFO 'Name : %', v_ename;
RAISE INFO 'Job : %', v_job;
RAISE INFO 'Manager : %', v_mgr;
RAISE INFO 'Hire Date : %', v_hiredate;
RAISE INFO 'Salary : %', v_sal;
RAISE INFO 'Commission : %', v_comm;
RETURN v_empno;
EXCEPTION
WHEN OTHERS THEN
RAISE INFO 'The following is SQLERRM : %', SQLERRM;
RAISE INFO 'The following is SQLSTATE: %', SQLSTATE;
RETURN -1;
END;
$$ LANGUAGE 'plpgsql';
--
-- Rule to INSERT into view 'salesemp'
--
CREATE OR REPLACE RULE salesemp_i AS ON INSERT TO salesemp
DO INSTEAD
INSERT INTO emp VALUES (NEW.empno, NEW.ename, 'SALESMAN', 7698,
NEW.hiredate, NEW.sal, NEW.comm, 30);
--
-- Rule to UPDATE view 'salesemp'
--
CREATE OR REPLACE RULE salesemp_u AS ON UPDATE TO salesemp
DO INSTEAD
UPDATE emp SET empno = NEW.empno,
ename = NEW.ename,
hiredate = NEW.hiredate,
sal = NEW.sal,
comm = NEW.comm
WHERE empno = OLD.empno;
--
-- Rule to DELETE from view 'salesemp'
--
CREATE OR REPLACE RULE salesemp_d AS ON DELETE TO salesemp
DO INSTEAD
DELETE FROM emp WHERE empno = OLD.empno;
--
-- After statement-level trigger that displays a message after
-- an insert, update, or deletion to the 'emp' table. One message
-- per SQL command is displayed.
--
CREATE OR REPLACE FUNCTION user_audit_trig() RETURNS TRIGGER
AS $$
DECLARE
v_action VARCHAR(24);
v_text TEXT;
BEGIN
IF TG_OP = 'INSERT' THEN
v_action := ' added employee(s) on ';
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 20
 EDB Postgres Advanced Server Guide

ELSIF TG_OP = 'UPDATE' THEN

v_action := ' updated employee(s) on ';
ELSIF TG_OP = 'DELETE' THEN
v_action := ' deleted employee(s) on ';
END IF;
v_text := 'User ' || USER || v_action || CURRENT_DATE;
RAISE INFO ' %', v_text;
RETURN NULL;
END;
$$ LANGUAGE 'plpgsql';
CREATE TRIGGER user_audit_trig
AFTER INSERT OR UPDATE OR DELETE ON emp
FOR EACH STATEMENT EXECUTE PROCEDURE user_audit_trig();
--
-- Before row-level trigger that displays employee number and
-- salary of an employee that is about to be added, updated,
-- or deleted in the 'emp' table.
--
CREATE OR REPLACE FUNCTION emp_sal_trig() RETURNS TRIGGER
AS $$
DECLARE
sal_diff NUMERIC(7,2);
BEGIN
IF TG_OP = 'INSERT' THEN
RAISE INFO 'Inserting employee %', NEW.empno;
RAISE INFO '..New salary: %', NEW.sal;
RETURN NEW;
END IF;
IF TG_OP = 'UPDATE' THEN
sal_diff := NEW.sal - OLD.sal;
RAISE INFO 'Updating employee %', OLD.empno;
RAISE INFO '..Old salary: %', OLD.sal;
RAISE INFO '..New salary: %', NEW.sal;
RAISE INFO '..Raise : %', sal_diff;
RETURN NEW;
END IF;
IF TG_OP = 'DELETE' THEN
RAISE INFO 'Deleting employee %', OLD.empno;
RAISE INFO '..Old salary: %', OLD.sal;
RETURN OLD;
END IF;
END;
$$ LANGUAGE 'plpgsql';
CREATE TRIGGER emp_sal_trig
BEFORE DELETE OR INSERT OR UPDATE ON emp
FOR EACH ROW EXECUTE PROCEDURE emp_sal_trig();
COMMIT;

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 21

 EDB Postgres Advanced Server Guide

2 Enhanced Compatibility Features

Advanced Server includes extended functionality that provides compatibility for syntax
supported by Oracle applications. Detailed information about all of the compatibility
features supported by Advanced Server is provided in the Database Compatibility for
Oracle Developers Guides; the information is broken into four guides:

The Database Compatibility for Oracle Developers Guide provides an overview

of the compatible procedural language, profile management, partitioning syntax,
and sample applications supported by Advanced Server.

The Database Compatibility for Oracle Developers Tools and Utilities Guide
provides information about the compatible tools supported by Advanced Server:
EDB*Plus, EDB*Loader, EDB*Wrap, and DRITA.

The Database Compatibility for Oracle Developers Built-in Packages Guide

provides information about using the compatible syntax available in the built-in
packages.

The Database Compatibility for Oracle Developers Reference Guide provides

reference information about using Advanced Server compatibility features,
including SQL syntax, compatible views and system tables, and data types.

Version-specific guides are available at:

https://www.enterprisedb.com/resources/product-documentation

The following sections highlight some of the compatibility features supported by

Advanced Server.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 22

 EDB Postgres Advanced Server Guide

2.1 Enabling Compatibility Features

There are several ways to install Advanced Server that will allow you to take advantage
of compatibility features:

Use the interactive graphical installer; on the Advanced Server Dialect dialog, use
the drop-down menu to select Compatible with Oracle.

Use the INITDBOPTS variable (in the Advanced Server service configuration file)
to specify --redwood-like before initializing your cluster.

When invoking initdb to initialize your cluster, include the --redwood-like

option.

For more information about the installation options supported by the Advanced Server
installers, please see the EDB Postgres Advanced Server Installation Guide, available
from the EDB website at:

https://www.enterprisedb.com/resources/product-documentation

2.2 Stored Procedural Language

Advanced Server supports a highly productive procedural language that allows you to
write custom procedures, functions, triggers and packages. The procedural language:

complements the SQL language and built-in packages.

provides a seamless development and testing environment.

allows you to create reusable code.

For information about using the Stored Procedural Language, see the Database
Compatibility for Oracle Developers Guide, available at:

https://www.enterprisedb.com/resources/product-documentation

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 23

 EDB Postgres Advanced Server Guide

2.3 Optimizer Hints

When you invoke a DELETE, INSERT, SELECT, or UPDATE command, the server
generates a set of execution plans; after analyzing those execution plans, the server
selects a plan that will (generally) return the result set in the least amount of time. The
server's choice of plan is dependent upon several factors:

The estimated execution cost of data handling operations.

Parameter values assigned to parameters in the Query Tuning section of the

postgresql.conf file.

Column statistics that have been gathered by the ANALYZE command.

As a rule, the query planner will select the least expensive plan. You can use an optimizer
hint to influence the server as it selects a query plan.

An optimizer hint is a directive (or multiple directives) embedded in a comment-like

syntax that immediately follows a DELETE, INSERT, SELECT or UPDATE command.
Keywords in the comment instruct the server to employ or avoid a specific plan when
producing the result set. For information about using optimizer hints, see the Database
Compatibility for Oracle Developers Guide, available at:

https://www.enterprisedb.com/resources/product-documentation

2.4 Data Dictionary Views

Advanced Server includes a set of views that provide information about database objects
in a manner compatible with the Oracle data dictionary views. For detailed information
about the views available with Advanced Server, please see the Database Compatibility
for Oracle Developers Reference Guide, available at:

https://www.enterprisedb.com/resources/product-documentation

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 24

 EDB Postgres Advanced Server Guide

2.5 dblink_ora

dblink_ora provides an OCI-based database link that allows you to SELECT, INSERT,
UPDATE or DELETE data stored on an Oracle system from within Advanced Server. For
detailed information about using dblink_ora, and the supported functions and
procedures, see the Database Compatibility for Oracle Developers Guide, available at:

https://www.enterprisedb.com/resources/product-documentation

2.6 Profile Management

Advanced Server supports compatible SQL syntax for profile management. Profile
management commands allow a database superuser to create and manage named
profiles. Each profile defines rules for password management that augment password
and md5 authentication. The rules in a profile can:

count failed login attempts

lock an account due to excessive failed login attempts
mark a password for expiration
define a grace period after a password expiration
define rules for password complexity
define rules that limit password re-use

A profile is a named set of attributes that allow you to easily manage a group of roles that
share comparable authentication requirements. If password requirements change, you
can modify the profile to have the new requirements applied to each user that is
associated with that profile.

After creating the profile, you can associate the profile with one or more users. When a
user connects to the server, the server enforces the profile that is associated with their
login role. Profiles are shared by all databases within a cluster, but each cluster may have
multiple profiles. A single user with access to multiple databases will use the same
profile when connecting to each database within the cluster.

For information about using profile management commands, see the Database
Compatibility for Oracle Developers Guide, available at:

https://www.enterprisedb.com/resources/product-documentation

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 25

 EDB Postgres Advanced Server Guide

2.7 Built-In Packages

Advanced Server supports a number of built-in packages that provide compatibility with
Oracle procedures and functions.

Package Name Description

DBMS_ALERT The DBMS_ALERT package provides the capability to register for, send,
and receive alerts.
DBMS_AQ The DBMS_AQ package provides message queueing and processing for
Advanced Server.
DBMS_AQADM The DBMS_AQADM package provides supporting procedures for
Advanced Queueing functionality.
DBMS_CRYPTO The DBMS_CRYPTO package provides functions and procedures that
allow you to encrypt or decrypt RAW, BLOB or CLOB data. You can also
use DBMS_CRYPTO functions to generate cryptographically strong
random values.
DBMS_JOB The DBMS_JOB package provides for the creation, scheduling, and
managing of jobs.
DBMS_LOB The DBMS_LOB package provides the capability to operate on large
objects.
DBMS_LOCK Advanced Server provides support for the DBMS_LOCK.SLEEP
procedure.
DBMS_MVIEW Use procedures in the DBMS_MVIEW package to manage and refresh
materialized views and their dependencies.
DBMS_OUTPUT The DBMS_OUTPUT package provides the capability to send messages to
a message buffer, or get messages from the message buffer.
DBMS_PIPE The DBMS_PIPE package provides the capability to send messages
through a pipe within or between sessions connected to the same database
cluster.
DBMS_PROFILER The DBMS_PROFILER package collects and stores performance
information about the PL/pgSQL and SPL statements that are executed
during a performance profiling session.
DBMS_RANDOM The DBMS_RANDOM package provides a number of methods to generate
random values. The procedures and functions available in the
DBMS_RANDOM package are listed in the following table.
DBMS_RLS The DBMS_RLS package enables the implementation of Virtual Private
Database on certain Advanced Server database objects.
DBMS_SCHEDULER The DBMS_SCHEDULER package provides a way to create and manage
jobs, programs and job schedules.
DBMS_SESSION Advanced Server provides support for the DBMS_SESSION.SET_ROLE
procedure.
DBMS_SQL The DBMS_SQL package provides an application interface to the
EnterpriseDB dynamic SQL functionality.
DBMS_UTILITY The DBMS_UTILITY package provides various utility programs.
UTL_ENCODE The UTL_ENCODE package provides a way to encode and decode data.
UTL_FILE The UTL_FILE package provides the capability to read from, and write to
files on the operating systems file system.
UTL_HTTP The UTL_HTTP package provides a way to use the HTTP or HTTPS
protocol to retrieve information found at an URL.
UTL_MAIL The UTL_MAIL package provides the capability to manage e-mail.
UTL_RAW The UTL_RAW package allows you to manipulate or retrieve the length of
raw data types.
UTL_SMTP The UTL_SMTP package provides the capability to send e-mails over the
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 26
 EDB Postgres Advanced Server Guide

Simple Mail Transfer Protocol (SMTP).

UTL_URL The UTL_URL package provides a way to escape illegal and reserved
characters within an URL.

For detailed information about the procedures and functions available within each
package, please see the Database Compatibility for Oracle Developers Built-In Package
Guide, available at:

https://www.enterprisedb.com/resources/product-documentation

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 27

 EDB Postgres Advanced Server Guide

2.8 Open Client Library

The Open Client Library provides application interoperability with the Oracle Call
Interface an application that was formerly locked in can now work with either an
Advanced Server or an Oracle database with minimal to no changes to the application
code. The EnterpriseDB implementation of the Open Client Library is written in C.

The following diagram compares the Open Client Library and Oracle Call Interface
application stacks.

Figure 2.1 The Open Client Library.

For detailed information about the functions supported by the Open Client Library, see
the EDB Postgres Advanced Server OCI Connector Guide, available at:

https://www.enterprisedb.com/resources/product-documentation

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 28

 EDB Postgres Advanced Server Guide

2.9 Utilities
For detailed information about the compatible syntax supported by the utilities listed
below, see the Database Compatibility for Oracle Developers Tools and Utilities Guide,
available at:

https://www.enterprisedb.com/resources/product-documentation

EDB*Plus

EDB*Plus is a utility program that provides a command line user interface to the
Advanced Server that will be familiar to Oracle developers and users. EDB*Plus accepts
SQL commands, SPL anonymous blocks, and EDB*Plus commands.

EDB*Plus allows you to:

Query certain database objects

Execute stored procedures
Format output from SQL commands
Execute batch scripts
Execute OS commands
Record output

For detailed information about EDB*Plus, please see the EDB*Plus User's Guide.

EDB*Loader

EDB*Loader is a high-performance bulk data loader that provides an interface

compatible with Oracle databases for Advanced Server. The EDB*Loader command line
utility loads data from an input source, typically a file, into one or more tables using a
subset of the parameters offered by Oracle SQL*Loader.

EDB*Loader features include:

Support for the Oracle SQL*Loader data loading methods - conventional path
load, direct path load, and parallel direct path load
Oracle SQL*Loader compatible syntax for control file directives
Input data with delimiter-separated or fixed-width fields
Bad file for collecting rejected records
Loading of multiple target tables
Discard file for collecting records that do not meet the selection criteria of any
target table
Log file for recording the EDB*Loader session and any error messages
Data loading from standard input and remote loading

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 29

 EDB Postgres Advanced Server Guide

EDB*Wrap

The EDB*Wrap utility protects proprietary source code and programs (functions, stored
procedures, triggers, and packages) from unauthorized scrutiny. The EDB*Wrap
program translates a file that contains SPL or PL/pgSQL source code (the plaintext) into
a file that contains the same code in a form that is nearly impossible to read. Once you
have the obfuscated form of the code, you can send that code to Advanced Server and it
will store those programs in obfuscated form. While EDB*Wrap does obscure code,
table definitions are still exposed.

Everything you wrap is stored in obfuscated form. If you wrap an entire package, the
package body source, as well as the prototypes contained in the package header and the
functions and procedures contained in the package body are stored in obfuscated form.

Dynamic Runtime Instrumentation Tools Architecture (DRITA)

The Dynamic Runtime Instrumentation Tools Architecture (DRITA) allows a DBA to

query catalog views to determine the wait events that affect the performance of individual
sessions or the system as a whole. DRITA records the number of times each event occurs
as well as the time spent waiting; you can use this information to diagnose performance
problems. DRITA offers this functionality, while consuming minimal system resources.

DRITA compares snapshots to evaluate the performance of a system. A snapshot is a

saved set of system performance data at a given point in time. Each snapshot is identified
by a unique ID number; you can use snapshot ID numbers with DRITA reporting
functions to return system performance statistics.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 30

 EDB Postgres Advanced Server Guide

2.10 ECPGPlus
EnterpriseDB has enhanced ECPG (the PostgreSQL pre-compiler) to create ECPGPlus.
ECPGPlus allows you to include embedded SQL commands in C applications; when you
use ECPGPlus to compile an application that contains embedded SQL commands, the
SQL code is syntax-checked and translated into C.

ECPGPlus supports Pro*C syntax in C programs when connected to an Advanced Server

database. ECPGPlus supports:

Oracle Dynamic SQL Method 4 (ODS-M4)

Pro*C compatible anonymous blocks
A CALL statement compatible with Oracle databases

For information about using ECPGPlus, please see the EDB Postgres Advanced Server
ECPG Connector Guide, available from the EnterpriseDB website at:

https://www.enterprisedb.com/resources/product-documentation

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 31

 EDB Postgres Advanced Server Guide

2.11 Table Partitioning

In a partitioned table, one logically large table is broken into smaller physical pieces.
Partitioning can provide several benefits:

Query performance can be improved dramatically in certain situations,

particularly when most of the heavily accessed rows of the table are in a single
partition or a small number of partitions. Partitioning allows you to omit the
partition column from the front of an index, reducing index size and making it
more likely that the heavily used parts of the index fits in memory.

When a query or update accesses a large percentage of a single partition,

performance may improve because the server will perform a sequential scan of
the partition instead of using an index and random access reads scattered across
the whole table.

A bulk load (or unload) can be implemented by adding or removing partitions, if

you plan that requirement into the partitioning design. ALTER TABLE is far faster
than a bulk operation. It also entirely avoids the VACUUM overhead caused by a
bulk DELETE.

Seldom-used data can be migrated to less-expensive (or slower) storage media.

Table partitioning is worthwhile only when a table would otherwise be very large. The
exact point at which a table will benefit from partitioning depends on the application; a
good rule of thumb is that the size of the table should exceed the physical memory of the
database server.

For information about database compatibility features supported by Advanced Server see
the Database Compatibility for Oracle Developer's Guide, available at:

https://www.enterprisedb.com/resources/product-documentation

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 32

 EDB Postgres Advanced Server Guide

3 Database Administration
This chapter describes the features that aid in the management and administration of
Advanced Server databases.

3.1 Configuration Parameters

This section describes the database server configuration parameters of Advanced Server.
These parameters control various aspects of the database servers behavior and
environment such as data file and log file locations, connection, authentication, and
security settings, resource allocation and consumption, archiving and replication settings,
error logging and statistics gathering, optimization and performance tuning, locale and
formatting settings, and so on.

Most of these configuration parameters apply to PostgreSQL as well. Configuration

parameters that apply only to Advanced Server are noted in Section 3.1.2.

Additional information about configuration parameters can be found in the PostgreSQL

Core Documentation available at:

https://www.postgresql.org/docs/10/static/runtime-config.html

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 33

 EDB Postgres Advanced Server Guide

3.1.1 Setting Configuration Parameters

This section provides an overview of how configuration parameters are specified and set.

Each configuration parameter is set using a name/value pair. Parameter names are case-
insensitive. The parameter name is typically separated from its value by an optional
equals sign (=).

The following is an example of some configuration parameter settings in the

postgresql.conf file:

# This is a comment
log_connections = yes
log_destination = 'syslog'
search_path = '"$user", public'
shared_buffers = 128MB

Parameter values are specified as one of five types:

Boolean. Acceptable values can be written as on, off, true, false, yes, no, 1,
0, or any unambiguous prefix of these.
Integer. Number without a fractional part.
Floating Point. Number with an optional fractional part separated by a decimal
point.
String. Text value. Enclose in single quotes if the value is not a simple identifier
or number (that is, the value contains special characters such as spaces or other
punctuation marks).
Enum. Specific set of string values. The allowed values can be found in the
system view pg_settings.enumvals. Enum values are case-insensitive.

Some settings specify a memory or time value. Each of these has an implicit unit, which
is kilobytes, blocks (typically 8 kilobytes), milliseconds, seconds, or minutes. Default
units can be found by referencing the system view pg_settings.unit. A different unit
can be specified explicitly.

Valid memory units are kB (kilobytes), MB (megabytes), and GB (gigabytes). Valid time
units are ms (milliseconds), s (seconds), min (minutes), h (hours), and d (days). The
multiplier for memory units is 1024.

The configuration parameter settings can be established in a number of different ways:

There is a number of parameter settings that are established when the Advanced
Server database product is built. These are read-only parameters, and their values
cannot be changed. There are also a couple of parameters that are permanently set

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 34

 EDB Postgres Advanced Server Guide

for each database when the database is created. These parameters are read-only as
well and cannot be subsequently changed for the database.
The initial settings for almost all configurable parameters across the entire
database cluster are listed in the configuration file, postgresql.conf. These
settings are put into effect upon database server start or restart. Some of these
initial parameter settings can be overridden as discussed in the following bullet
points. All configuration parameters have built-in default settings that are in effect
if not explicitly overridden.
Configuration parameters in the postgresql.conf file are overridden when the
same parameters are included in the postgresql.auto.conf file. The ALTER
SYSTEM command is used to manage the configuration parameters in the
postgresql.auto.conf file.
Parameter settings can be modified in the configuration file while the database
server is running. If the configuration file is then reloaded (meaning a SIGHUP
signal is issued), for certain parameter types, the changed parameters settings
immediately take effect. For some of these parameter types, the new settings are
available in a currently running session immediately after the reload. For other of
these parameter types, a new session must be started to use the new settings. And
yet for other parameter types, modified settings do not take effect until the
database server is stopped and restarted. See Section 18.1, Setting Parameters in
the PostgreSQL Core Documentation for information on how to reload the
configuration file.
The SQL commands ALTER DATABASE, ALTER ROLE, or ALTER ROLE IN
DATABASE can be used to modify certain parameter settings. The modified
parameter settings take effect for new sessions after the command is executed.
ALTER DATABASE affects new sessions connecting to the specified database.
ALTER ROLE affects new sessions started by the specified role. ALTER ROLE IN
DATABASE affects new sessions started by the specified role connecting to the
specified database. Parameter settings established by these SQL commands
remain in effect indefinitely, across database server restarts, overriding settings
established by the methods discussed in the second and third bullet points.
Parameter settings established using the ALTER DATABASE, ALTER ROLE, or
ALTER ROLE IN DATABASE commands can only be changed by: a) re-issuing
these commands with a different parameter value, or b) issuing these commands
using either of the SET parameter TO DEFAULT clause or the RESET
parameter clause. These clauses change the parameter back to using the setting
established by the methods set forth in the prior bullet points. See Section I, SQL
Commands of Chapter VI Reference in the PostgreSQL Core Documentation
for the exact syntax of these SQL commands.
Changes can be made for certain parameter settings for the duration of individual
sessions using the PGOPTIONS environment variable or by using the SET
command within the EDB-PSQL or PSQL command line terminal programs.
Parameter settings made in this manner override settings established using any of
the methods described by the second, third, and fourth bullet points, but only for
the duration of the session.
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 35
 EDB Postgres Advanced Server Guide

3.1.2 Summary of Configuration Parameters

This section contains a summary table listing all Advanced Server configuration
parameters along with a number of key attributes of the parameters.

These attributes are described by the following columns of the summary table:

Parameter. Configuration parameter name.

Scope of Effect. Scope of effect of the configuration parameter setting. Cluster
Setting affects the entire database cluster (that is, all databases managed by the
database server instance). Database Setting can vary by database and is
established when the database is created. Applies to a small number of parameters
related to locale settings. Session Setting can vary down to the granularity of
individual sessions. In other words, different settings can be made for the
following entities whereby the latter settings in this list override prior ones: a) the
entire database cluster, b) specific databases in the database cluster, c) specific
roles, d) specific roles when connected to specific databases, e) a specific session.
When Takes Effect. When a changed parameter setting takes effect. Preset
Established when the Advanced Server product is built or a particular database is
created. This is a read-only parameter and cannot be changed. Restart
Database server must be restarted. Reload Configuration file must be reloaded
(or the database server can be restarted). Immediate Immediately effective in a
session if the PGOPTIONS environment variable or the SET command is used to
change the setting in the current session. Effective in new sessions if ALTER
DATABASE, ALTER ROLE, or ALTER ROLE IN DATABASE commands are used
to change the setting.
Authorized User. Type of operating system account or database role that must be
used to put the parameter setting into effect. EPAS service account EDB
Postgres Advanced Server service account (enterprisedb for an installation
compatible with Oracle databases, postgres for a PostgreSQL compatible mode
installation). Superuser Database role with superuser privileges. User Any
database role with permissions on the affected database object (the database or
role to be altered with the ALTER command). n/a Parameter setting cannot be
changed by any user.
Description. Brief description of the configuration parameter.
EPAS Only. X Configuration parameter is applicable to EDB Postgres
Advanced Server only. No entry in this column indicates the configuration
parameter applies to PostgreSQL as well.

Note: There are a number of parameters that should never be altered. These are
designated as Note: For internal use only in the Description column.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 36

 EDB Postgres Advanced Server Guide

Table 3-1 - Summary of Configuration Parameters

When EPA
Scope of Authorized
Parameter Takes Description S
Effect User
Effect Only
allow_system_table_mods Cluster Restart EPAS Allows modifications of the
service structure of system tables.
account
application_name Session Immediate User Sets the application name to be
reported in statistics and logs.
archive_command Cluster Reload EPAS Sets the shell command that
service will be called to archive a WAL
account file.
archive_mode Cluster Restart EPAS Allows archiving of WAL files
service using archive_command.
account
archive_timeout Cluster Reload EPAS Forces a switch to the next xlog
service file if a new file has not been
account started within N seconds.
array_nulls Session Immediate User Enable input of NULL elements
in arrays.
authentication_timeout Cluster Reload EPAS Sets the maximum allowed time
service to complete client
account authentication.
autovacuum Cluster Reload EPAS Starts the autovacuum
service subprocess.
account
autovacuum_analyze_scale Cluster Reload EPAS Number of tuple inserts,
_factor service updates or deletes prior to
account analyze as a fraction of
reltuples.
autovacuum_analyze_thres Cluster Reload EPAS Minimum number of tuple
hold service inserts, updates or deletes prior
account to analyze.
autovacuum_freeze_max_ag Cluster Restart EPAS Age at which to autovacuum a
e service table to prevent transaction ID
account wraparound.
autovacuum_max_workers Cluster Restart EPAS Sets the maximum number of
service simultaneously running
account autovacuum worker processes.
autovacuum_multixact_fre Cluster Restart EPAS Multixact age at which to
eze_max_age service autovacuum a table to prevent
account multixact wraparound.
autovacuum_naptime Cluster Reload EPAS Time to sleep between
service autovacuum runs.
account
autovacuum_vacuum_cost_d Cluster Reload EPAS Vacuum cost delay in
elay service milliseconds, for autovacuum.
account
autovacuum_vacuum_cost_l Cluster Reload EPAS Vacuum cost amount available
imit service before napping, for
account autovacuum.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 37

 EDB Postgres Advanced Server Guide

When EPA
Scope of Authorized
Parameter Takes Description S
Effect User
Effect Only
autovacuum_vacuum_scale_ Cluster Reload EPAS Number of tuple updates or
factor service deletes prior to vacuum as a
account fraction of reltuples.

autovacuum_vacuum_thresh Cluster Reload EPAS Minimum number of tuple

old service updates or deletes prior to
account vacuum.
autovacuum_work_mem Cluster Reload EPAS Sets the maximum memory to
service be used by each autovacuum
account worker process.
backslash_quote Session Immediate User Sets whether "\'" is allowed in
string literals.
bgwriter_delay Cluster Reload EPAS Background writer sleep time
service between rounds.
account
bgwriter_lru_maxpages Cluster Reload EPAS Background writer maximum
service number of LRU pages to flush
account per round.
bgwriter_lru_multiplier Cluster Reload EPAS Multiple of the average buffer
service usage to free per round.
account
block_size Cluster Preset n/a Shows the size of a disk block.
bonjour Cluster Restart EPAS Enables advertising the server
service via Bonjour.
account
bonjour_name Cluster Restart EPAS Sets the Bonjour service name.
service
account
bytea_output Session Immediate User
Sets the output format for
bytea.
check_function_bodies Session Immediate User Check function bodies during
CREATE FUNCTION.
checkpoint_completion_ta Cluster Reload EPAS Time spent flushing dirty
rget service buffers during checkpoint, as
account fraction of checkpoint interval.
checkpoint_segments Deprecated in Deprecated Deprecated This parameter is not supported
9.5 in 9.5 in 9.5 by server version 9.5 or later.
Specifying a value for the
parameter will prevent the
server from starting.
checkpoint_timeout Cluster Reload EPAS Sets the maximum time
service between automatic WAL
account checkpoints.
checkpoint_warning Cluster Reload EPAS Enables warnings if checkpoint
service segments are filled more
account frequently than this.
client_encoding Session Immediate User Sets the client's character set
encoding.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 38

 EDB Postgres Advanced Server Guide

When EPA
Scope of Authorized
Parameter Takes Description S
Effect User
Effect Only
client_min_messages Session Immediate User Sets the message levels that are
sent to the client.
commit_delay Session Immediate Superuser Sets the delay in microseconds
between transaction commit
and flushing WAL to disk.
commit_siblings Session Immediate User Sets the minimum concurrent
open transactions before
performing commit_delay.
config_file Cluster Restart EPAS Sets the server's main
service configuration file.
account
constraint_exclusion Session Immediate User Enables the planner to use
constraints to optimize queries.
cpu_index_tuple_cost Session Immediate User Sets the planner's estimate of
the cost of processing each
index entry during an index
scan.
cpu_operator_cost Session Immediate User Sets the planner's estimate of
the cost of processing each
operator or function call.
cpu_tuple_cost Session Immediate User Sets the planner's estimate of
the cost of processing each
tuple (row).
cursor_tuple_fraction Session Immediate User Sets the planner's estimate of
the fraction of a cursor's rows
that will be retrieved.
custom_variable_classes Cluster Reload EPAS Deprecated in Advanced Server X
service 9.2.
account
data_checksums Cluster Preset n/a Shows whether data checksums
are turned on for this cluster.
data_directory Cluster Restart EPAS Sets the server's data directory.
service
account
DateStyle Session Immediate User Sets the display format for date
and time values.
db_dialect Session Immediate User Sets the precedence of built-in X
namespaces.
dbms_alert.max_alerts Cluster Restart EPAS Sets maximum number of X
service alerts.
account
dbms_pipe.total_message_ Cluster Restart EPAS Specifies the total size of the X
buffer service buffer used for the
account DBMS_PIPE package.
db_user_namespace Cluster Reload EPAS Enables per-database user
service names.
account
deadlock_timeout Session Immediate Superuser Sets the time to wait on a lock

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 39

 EDB Postgres Advanced Server Guide

When EPA
Scope of Authorized
Parameter Takes Description S
Effect User
Effect Only
before checking for deadlock.
debug_assertions Cluster Preset n/a Turns on various assertion
checks. (Not supported in
EPAS builds.)
debug_pretty_print Session Immediate User Indents parse and plan tree
displays.
debug_print_parse Session Immediate User Logs each query's parse tree.
debug_print_plan Session Immediate User Logs each query's execution
plan.
debug_print_rewritten Session Immediate User Logs each query's rewritten
parse tree.
default_heap_fillfactor Session Immediate User Create new tables with this X
heap fillfactor by default.
default_statistics_targe Session Immediate User Sets the default statistics target.
t

default_tablespace Session Immediate User Sets the default tablespace to

create tables and indexes in.
default_text_search_conf Session Immediate User Sets default text search
ig configuration.
default_transaction_defe Session Immediate User Sets the default deferrable
rrable status of new transactions.
default_transaction_isol Session Immediate User Sets the transaction isolation
ation level of each new transaction.
default_transaction_read Session Immediate User Sets the default read-only status
_only of new transactions.
default_with_oids Session Immediate User Create new tables with OIDs by
default.
default_with_rowids Session Immediate User Create new tables with ROWID X
support (OIDs with indexes) by
default.
dynamic_library_path Session Immediate Superuser Sets the path for dynamically
loadable modules.
dynamic_shared_memory_ty Cluster Restart EPAS Selects the dynamic shared
pe service memory implementation used.
account
edb_audit Cluster Reload EPAS Enable EDB Auditing to create X
service audit reports in XML or CSV
account format.
edb_audit_connect Cluster Reload EPAS Audits each successful X
service connection.
account
edb_audit_destination Cluster Reload EPAS Sets edb_audit_directory X
service or syslog as the destination
account directory for audit files.
edb_audit_directory Cluster Reload EPAS Sets the destination directory X
service for audit files.
account
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 40
 EDB Postgres Advanced Server Guide

When EPA
Scope of Authorized
Parameter Takes Description S
Effect User
Effect Only
edb_audit_disconnect Cluster Reload EPAS Audits end of a session. X
service
account
edb_audit_filename Cluster Reload EPAS Sets the file name pattern for X
service audit files.
account
edb_audit_rotation_day Cluster Reload EPAS Automatic rotation of log files X
service based on day of week.
account
edb_audit_rotation_secon Cluster Reload EPAS Automatic log file rotation will X
ds service occur after N seconds.
account
edb_audit_rotation_size Cluster Reload EPAS Automatic log file rotation will X
service occur after N Megabytes.
account
edb_audit_statement Cluster Reload EPAS Sets the type of statements to X
service audit.
account
edb_audit_tag Session Immediate User Specify a tag to be included in X
the audit log.
edb_connectby_order Session Immediate User Sort results of CONNECT BY X
queries with no ORDER BY to
depth-first order. Note: For
internal use only.
edb_custom_plan_tries Session Immediate User Specifies the number of custom X
execution plans considered by
the planner before the planner
selects a generic execution
plan.
edb_dynatune Cluster Restart EPAS Sets the edb utilization X
service percentage.
account
edb_dynatune_profile Cluster Restart EPAS Sets the workload profile for X
service dynatune.
account
edb_enable_icache Cluster Restart EPAS Enable external shared buffer X
service Infinite Cache mechanism.
account
edb_enable_pruning Session Immediate User Enables the planner to early- X
prune partitioned tables.
edb_icache_compression_l Session Immediate Superuser Sets compression level of X
evel Infinite Cache buffers.
edb_icache_servers Cluster Reload EPAS A list of comma separated X
service hostname:portnumber
account icache servers.
edb_log_every_bulk_value Session Immediate Superuser Sets the statements logged for X
bulk processing.
edb_max_resource_groups Cluster Restart EPAS Specifies the maximum number X

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 41

 EDB Postgres Advanced Server Guide

When EPA
Scope of Authorized
Parameter Takes Description S
Effect User
Effect Only
service of resource groups for
account simultaneous use.
edb_max_spins_per_delay Cluster Restart EPAS Specifies the number of times a X
service session will spin while waiting
account for a lock.
edb_redwood_date Session Immediate User Determines whether DATE X
should behave like a
TIMESTAMP or not.
edb_redwood_greatest_lea Session Immediate User Determines how GREATEST and X
st LEAST functions should handle
NULL parameters.
edb_redwood_raw_names Session Immediate User Return the unmodified name X
stored in the PostgreSQL
system catalogs from Redwood
interfaces.
edb_redwood_strings Session Immediate User Treat NULL as an empty string X
when concatenated with a text
value.
edb_resource_group Session Immediate User Specifies the resource group to X
be used by the current process.
edb_sql_protect.enabled Cluster Reload EPAS Defines whether SQL/Protect X
service should track queries or not.
account
edb_sql_protect.level Cluster Reload EPAS Defines the behavior of X
service SQL/Protect when an event is
account found.
edb_sql_protect.max_prot Cluster Restart EPAS Sets the maximum number of X
ected_relations service relations protected by
account SQL/Protect per role.
edb_sql_protect.max_prot Cluster Restart EPAS Sets the maximum number of X
ected_roles service roles protected by SQL/Protect.
account
edb_sql_protect.max_quer Cluster Restart EPAS Sets the maximum number of X
ies_to_save service offending queries to save by
account SQL/Protect.
edb_stmt_level_tx Session Immediate User Allows continuing on errors X
instead of requiring a
transaction abort.
edbldr.empty_csv_field Session Immediate Superuser Specifies how EDB*Loader X
handles empty strings.
effective_cache_size Session Immediate User Sets the planner's assumption
about the size of the disk cache.
effective_io_concurrency Session Immediate User Number of simultaneous
requests that can be handled
efficiently by the disk
subsystem.
enable_bitmapscan Session Immediate User Enables the planner's use of
bitmap-scan plans.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 42

 EDB Postgres Advanced Server Guide

When EPA
Scope of Authorized
Parameter Takes Description S
Effect User
Effect Only
enable_hashagg Session Immediate User Enables the planner's use of
hashed aggregation plans.
enable_hashjoin Session Immediate User Enables the planner's use of
hash join plans.
enable_hints Session Immediate User Enable optimizer hints in SQL X
statements.
enable_indexonlyscan Session Immediate User Enables the planners use of
index-only-scan plans.
enable_indexscan Session Immediate User Enables the planner's use of
index-scan plans.
enable_material Session Immediate User Enables the planner's use of
materialization.
enable_mergejoin Session Immediate User Enables the planner's use of
merge join plans.
enable_nestloop Session Immediate User Enables the planner's use of
nested-loop join plans.
enable_seqscan Session Immediate User Enables the planner's use of
sequential-scan plans.
enable_sort Session Immediate User Enables the planner's use of
explicit sort steps.
enable_tidscan Session Immediate User Enables the planner's use of
TID scan plans.
escape_string_warning Session Immediate User Warn about backslash escapes
in ordinary string literals.
event_source Cluster Restart EPAS Sets the application name used
service to identify PostgreSQL
account messages in the event log.
exit_on_error Session Immediate User Terminate session on any error.
external_pid_file Cluster Restart EPAS Writes the postmaster PID to
service the specified file.
account
extra_float_digits Session Immediate User Sets the number of digits
displayed for floating-point
values.
from_collapse_limit Session Immediate User Sets the FROM-list size beyond
which subqueries are not
collapsed.
fsync Cluster Reload EPAS Forces synchronization of
service updates to disk.
account
full_page_writes Cluster Reload EPAS Writes full pages to WAL when
service first modified after a
account checkpoint.
geqo Session Immediate User Enables genetic query
optimization.
geqo_effort Session Immediate User GEQO: effort is used to set the
default for other GEQO
parameters.
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 43
 EDB Postgres Advanced Server Guide

When EPA
Scope of Authorized
Parameter Takes Description S
Effect User
Effect Only
geqo_generations Session Immediate User GEQO: number of iterations of
the algorithm.
geqo_pool_size Session Immediate User GEQO: number of individuals
in the population.
geqo_seed Session Immediate User GEQO: seed for random path
selection.
geqo_selection_bias Session Immediate User GEQO: selective pressure
within the population.
geqo_threshold Session Immediate User Sets the threshold of FROM
items beyond which GEQO is
used.
gin_fuzzy_search_limit Session Immediate User Sets the maximum allowed
result for exact search by GIN.
hba_file Cluster Restart EPAS Sets the server's "hba"
service configuration file.
account
hot_standby Cluster Restart EPAS Allows connections and queries
service during recovery.
account
hot_standby_feedback Cluster Reload EPAS Allows feedback from a hot
service standby to the primary that will
account avoid query conflicts.
huge_pages Cluster Restart EPAS Use of huge pages on Linux.
service
account
icu_short_form Database Preset n/a Shows the ICU collation order X
configuration.
ident_file Cluster Restart EPAS Sets the server's "ident"
service configuration file.
account
ignore_checksum_failure Session Immediate Superuser Continues processing after a
checksum failure.
ignore_system_indexes Cluster/ Reload/ EPAS Disables reading from system
Session Immediate service indexes. (Can also be set with
account/ PGOPTIONS at session start.)
User
index_advisor.enabled Session Immediate User Enable Index Advisor plugin. X
integer_datetimes Cluster Preset n/a Datetimes are integer based.
IntervalStyle Session Immediate User Sets the display format for
interval values.
join_collapse_limit Session Immediate User Sets the FROM-list size beyond
which JOIN constructs are not
flattened.
krb_caseins_users Cluster Reload EPAS Sets whether Kerberos and
service GSSAPI user names should be
account treated as case-insensitive.
krb_server_keyfile Cluster Reload EPAS Sets the location of the
service Kerberos server key file.
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 44
 EDB Postgres Advanced Server Guide

When EPA
Scope of Authorized
Parameter Takes Description S
Effect User
Effect Only
account
lc_collate Database Preset n/a Shows the collation order
locale.
lc_ctype Database Preset n/a Shows the character
classification and case
conversion locale.
lc_messages Session Immediate Superuser Sets the language in which
messages are displayed.
lc_monetary Session Immediate User Sets the locale for formatting
monetary amounts.
lc_numeric Session Immediate User Sets the locale for formatting
numbers.
lc_time Session Immediate User Sets the locale for formatting
date and time values.
listen_addresses Cluster Restart EPAS Sets the host name or IP
service address(es) to listen to.
account
local_preload_libraries Cluster/ Reload/ EPAS Lists shared libraries to preload
Session Immediate service into each backend. (Can also be
account/ set with PGOPTIONS at session
User start.)
lock_timeout Session Immediate User Sets the maximum time allowed
that a statement may wait for a
lock.
lo_compat_privileges Session Immediate Superuser Enables backward compatibility
mode for privilege checks on
large objects.
log_autovacuum_min_durat Cluster Reload EPAS Sets the minimum execution
ion service time above which autovacuum
account actions will be logged.
log_checkpoints Cluster Reload EPAS Logs each checkpoint.
service
account
log_connections Cluster/ Reload/ EPAS Logs each successful
Session Immediate service connection. (Can also be set
account/ with PGOPTIONS at session
User start.)
log_destination Cluster Reload EPAS Sets the destination for server
service log output.
account
log_directory Cluster Reload EPAS Sets the destination directory
service for log files.
account
log_disconnections Cluster/ Reload/ EPAS Logs end of a session, including
Session Immediate service duration. (Can also be set with
account/ PGOPTIONS at session start.)
User
log_duration Session Immediate Superuser Logs the duration of each

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 45

 EDB Postgres Advanced Server Guide

When EPA
Scope of Authorized
Parameter Takes Description S
Effect User
Effect Only
completed SQL statement.
log_error_verbosity Session Immediate Superuser Sets the verbosity of logged
messages.
log_executor_stats Session Immediate Superuser Writes executor performance
statistics to the server log.
log_file_mode Cluster Reload EPAS Sets the file permissions for log
service files.
account
log_filename Cluster Reload EPAS Sets the file name pattern for
service log files.
account
log_hostname Cluster Reload EPAS Logs the host name in the
service connection logs.
account
log_line_prefix Cluster Reload EPAS Controls information prefixed
service to each log line.
account
log_lock_waits Session Immediate Superuser Logs long lock waits.
log_min_duration_stateme Session Immediate Superuser Sets the minimum execution
nt time above which statements
will be logged.
log_min_error_statement Session Immediate Superuser Causes all statements
generating error at or above this
level to be logged.
log_min_messages Session Immediate Superuser Sets the message levels that are
logged.
log_parser_stats Session Immediate Superuser Writes parser performance
statistics to the server log.
log_planner_stats Session Immediate Superuser Writes planner performance
statistics to the server log.
log_rotation_age Cluster Reload EPAS Automatic log file rotation will
service occur after N minutes.
account
log_rotation_size Cluster Reload EPAS Automatic log file rotation will
service occur after N kilobytes.
account
log_statement Session Immediate Superuser Sets the type of statements
logged.
log_statement_stats Session Immediate Superuser Writes cumulative performance
statistics to the server log.
log_temp_files Session Immediate Superuser Log the use of temporary files
larger than this number of
kilobytes.
log_timezone Cluster Reload EPAS Sets the time zone to use in log
service messages.
account
log_truncate_on_rotation Cluster Reload EPAS Truncate existing log files of
service same name during log rotation.
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 46
 EDB Postgres Advanced Server Guide

When EPA
Scope of Authorized
Parameter Takes Description S
Effect User
Effect Only
account
logging_collector Cluster Restart EPAS Start a subprocess to capture
service stderr output and/or csvlogs
account into log files.
maintenance_work_mem Session Immediate User Sets the maximum memory to
be used for maintenance
operations.
max_connections Cluster Restart EPAS Sets the maximum number of
service concurrent connections.
account
max_files_per_process Cluster Restart EPAS Sets the maximum number of
service simultaneously open files for
account each server process.
max_function_args Cluster Preset n/a Shows the maximum number of
function arguments.
max_identifier_length Cluster Preset n/a Shows the maximum identifier
length.
max_index_keys Cluster Preset n/a Shows the maximum number of
index keys.
max_locks_per_transactio Cluster Restart EPAS Sets the maximum number of
n service locks per transaction.
account
max_pred_locks_per_trans Cluster Restart EPAS Sets the maximum number of
action service predicate locks per transaction.
account
max_prepared_transaction Cluster Restart EPAS Sets the maximum number of
s service simultaneously prepared
account transactions.
max_replication_slots Cluster Restart EPAS Sets the maximum number of
service simultaneously defined
account replication slots.
max_stack_depth Session Immediate Superuser Sets the maximum stack depth,
in kilobytes.
max_standby_archive_dela Cluster Reload EPAS Sets the maximum delay before
y service canceling queries when a hot
account standby server is processing
archived WAL data.
max_standby_streaming_de Cluster Reload EPAS Sets the maximum delay before
lay service canceling queries when a hot
account standby server is processing
streamed WAL data.
max_wal_senders Cluster Restart EPAS Sets the maximum number of
service simultaneously running WAL
account sender processes.
max_wal_size Cluster Reload EPAS Sets the maximum size to
service which the WAL will grow
account between automatic WAL
checkpoints. The default is

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 47

 EDB Postgres Advanced Server Guide

When EPA
Scope of Authorized
Parameter Takes Description S
Effect User
Effect Only
1GB.
max_worker_processes Cluster Restart EPAS Maximum number of
service concurrent worker processes.
account
min_wal_size Cluster Reload EPAS Sets the threshold at which
service WAL logs will be recycled
account rather than removed. The
default is 80 MB.
nls_length_semantics Session Immediate Superuser Sets the semantics to use for X
char, varchar, varchar2 and
long columns.
odbc_lib_path Cluster Restart EPAS Sets the path for ODBC library. X
service
account
optimizer_mode Session Immediate User Default optimizer mode. X
oracle_home Cluster Restart EPAS Sets the path for the Oracle X
service home directory.
account
password_encryption Session Immediate User Encrypt passwords.
pg_prewarm.autoprewarm Cluster Restart EPAS Enables the autoprewarm X
service background worker.
account
pg_prewarm.autoprewarm_i Cluster Reload EPAS Sets the minimum number of X
nterval service seconds after which
account autoprewarm dumps
shared buffers.
port Cluster Restart EPAS Sets the TCP port on which the
service server listens.
account
post_auth_delay Cluster/ Reload/ EPAS Waits N seconds on connection
Session Immediate service startup after authentication.
account/ (Can also be set with
User PGOPTIONS at session start.)
pre_auth_delay Cluster Reload EPAS Waits N seconds on connection
service startup before authentication.
account
qreplace_function Session Immediate Superuser The function to be used by X
Query Replace feature. Note:
For internal use only.
query_rewrite_enabled Session Immediate User Child table scans will be X
skipped if their constraints
guarantee that no rows match
the query.
query_rewrite_integrity Session Immediate Superuser Sets the degree to which query X
rewriting must be enforced.
quote_all_identifiers Session Immediate User When generating SQL
fragments, quote all identifiers.
random_page_cost Session Immediate User Sets the planner's estimate of

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 48

 EDB Postgres Advanced Server Guide

When EPA
Scope of Authorized
Parameter Takes Description S
Effect User
Effect Only
the cost of a nonsequentially
fetched disk page.
restart_after_crash Cluster Reload EPAS Reinitialize server after
service backend crash.
account
search_path Session Immediate User Sets the schema search order
for names that are not schema-
qualified.
segment_size Cluster Preset n/a Shows the number of pages per
disk file.
seq_page_cost Session Immediate User Sets the planner's estimate of
the cost of a sequentially
fetched disk page.
server_encoding Database Preset n/a Sets the server (database)
character set encoding.
server_version Cluster Preset n/a Shows the server version.
server_version_num Cluster Preset n/a Shows the server version as an
integer.
session_preload_librarie Session Immediate Superuser Lists shared libraries to preload
s but only at into each backend.
connection
start
session_replication_role Session Immediate Superuser Sets the session's behavior for
triggers and rewrite rules.
shared_buffers Cluster Restart EPAS Sets the number of shared
service memory buffers used by the
account server.
shared_preload_libraries Cluster Restart EPAS Lists shared libraries to preload
service into server.
account
sql_inheritance Session Immediate User Causes subtables to be included
by default in various
commands.
ssl Cluster Restart EPAS Enables SSL connections.
service
account
ssl_ca_file Cluster Restart EPAS Location of the SSL certificate
service authority file.
account
ssl_cert_file Cluster Restart EPAS Location of the SSL server
service certificate file.
account
ssl_ciphers Cluster Restart EPAS Sets the list of allowed SSL
service ciphers.
account
ssl_crl_file Cluster Restart EPAS Location of the SSL certificate
service revocation list file.
account

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 49

 EDB Postgres Advanced Server Guide

When EPA
Scope of Authorized
Parameter Takes Description S
Effect User
Effect Only
ssl_ecdh_curve Cluster Restart EPAS Sets the curve to use for ECDH.
service
account
ssl_key_file Cluster Restart EPAS Location of the SSL server
service private key file.
account
ssl_prefer_server_cipher Cluster Restart EPAS Give priority to server
s service ciphersuite order.
account
ssl_renegotiation_limit Session Immediate User Set the amount of traffic to send
and receive before
renegotiating the encryption
keys.
standard_conforming_stri Session Immediate User Causes '...' strings to treat
ngs backslashes literally.
statement_timeout Session Immediate User Sets the maximum allowed
duration of any statement.
stats_temp_directory Cluster Reload EPAS Writes temporary statistics files
service to the specified directory.
account
superuser_reserved_conne Cluster Restart EPAS Sets the number of connection
ctions service slots reserved for superusers.
account
synchronize_seqscans Session Immediate User Enable synchronized sequential
scans.
synchronous_commit Session Immediate User Sets immediate fsync at
commit.
synchronous_standby_name Cluster Reload EPAS List of names of potential
s service synchronous standbys.
account
syslog_facility Cluster Reload EPAS Sets the syslog "facility" to be
service used when syslog enabled.
account
syslog_ident Cluster Reload EPAS Sets the program name used to
service identify PostgreSQL messages
account in syslog.
tcp_keepalives_count Session Immediate User Maximum number of TCP
keepalive retransmits.
tcp_keepalives_idle Session Immediate User Time between issuing TCP
keepalives.
tcp_keepalives_interval Session Immediate User Time between TCP keepalive
retransmits.
temp_buffers Session Immediate User Sets the maximum number of
temporary buffers used by each
session.
temp_file_limit Session Immediate Superuser Limits the total size of all
temporary files used by each
session.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 50

 EDB Postgres Advanced Server Guide

When EPA
Scope of Authorized
Parameter Takes Description S
Effect User
Effect Only
temp_tablespaces Session Immediate User Sets the tablespace(s) to use for
temporary tables and sort files.
timed_statistics Session Immediate User Enables the recording of timed X
statistics.
timezone Session Immediate User Sets the time zone for
displaying and interpreting time
stamps.
timezone_abbreviations Session Immediate User Selects a file of time zone
abbreviations.
trace_hints Session Immediate User Emit debug info about hints X
being honored.
trace_notify Session Immediate User Generates debugging output for
LISTEN and NOTIFY.
trace_recovery_messages Cluster Reload EPAS Enables logging of recovery-
service related debugging information.
account
trace_sort Session Immediate User Emit information about
resource usage in sorting.
track_activities Session Immediate Superuser Collects information about
executing commands.
track_activity_query_siz Cluster Restart EPAS Sets the size reserved for
e service pg_stat_activity.curren
account t_query, in bytes.
track_counts Session Immediate Superuser Collects statistics on database
activity.
track_functions Session Immediate Superuser Collects function-level statistics
on database activity.
track_io_timing Session Immediate Superuser Collects timing statistics for
database I/O activity.
transaction_deferrable Session Immediate User Whether to defer a read-only
serializable transaction until it
can be executed with no
possible serialization failures.
transaction_isolation Session Immediate User Sets the current transaction's
isolation level.
transaction_read_only Session Immediate User Sets the current transaction's
read-only status.
transform_null_equals Session Immediate User Treats "expr=NULL" as "expr
IS NULL".
unix_socket_directories Cluster Restart EPAS Sets the directory where the
service Unix-domain socket will be
account created.
unix_socket_group Cluster Restart EPAS Sets the owning group of the
service Unix-domain socket.
account
unix_socket_permissions Cluster Restart EPAS Sets the access permissions of
service the Unix-domain socket.
account
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 51
 EDB Postgres Advanced Server Guide

When EPA
Scope of Authorized
Parameter Takes Description S
Effect User
Effect Only
update_process_title Session Immediate Superuser Updates the process title to
show the active SQL command.
utl_encode.uudecode_redw Session Immediate User Allows decoding of Oracle- X
ood created uuencoded data.
utl_file.umask Session Immediate User Umask used for files created X
through the UTL_FILE
package.
vacuum_cost_delay Session Immediate User Vacuum cost delay in
milliseconds.
vacuum_cost_limit Session Immediate User Vacuum cost amount available
before napping.
vacuum_cost_page_dirty Session Immediate User Vacuum cost for a page dirtied
by vacuum.
vacuum_cost_page_hit Session Immediate User Vacuum cost for a page found
in the buffer cache.
vacuum_cost_page_miss Session Immediate User Vacuum cost for a page not
found in the buffer cache.
vacuum_defer_cleanup_age Cluster Reload EPAS Number of transactions by
service which VACUUM and HOT
account cleanup should be deferred, if
any.
vacuum_freeze_min_age Session Immediate User Minimum age at which
VACUUM should freeze a
table row.
vacuum_freeze_table_age Session Immediate User Age at which VACUUM
should scan whole table to
freeze tuples.
vacuum_multixact_freeze_ Session Immediate User Minimum age at which
min_age VACUUM should freeze a
MultiXactId in a table row.
vacuum_multixact_freeze_ Session Immediate User Multixact age at which
table_age VACUUM should scan whole
table to freeze tuples.
wal_block_size Cluster Preset n/a Shows the block size in the
write ahead log.
wal_buffers Cluster Restart EPAS Sets the number of disk-page
service buffers in shared memory for
account WAL.
wal_keep_segments Cluster Reload EPAS Sets the number of WAL files
service held for standby servers.
account
wal_level Cluster Restart EPAS Set the level of information
service written to the WAL.
account
wal_log_hints Cluster Restart EPAS Writes full pages to WAL when
service first modified after a
account checkpoint, even for non-
critical modifications.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 52

 EDB Postgres Advanced Server Guide

When EPA
Scope of Authorized
Parameter Takes Description S
Effect User
Effect Only
wal_receiver_status_inte Cluster Reload EPAS Sets the maximum interval
rval service between WAL receiver status
account reports to the primary.
wal_receiver_timeout Cluster Reload EPAS Sets the maximum wait time to
service receive data from the primary.
account
wal_segment_size Cluster Preset n/a Shows the number of pages per
write ahead log segment.
wal_sender_timeout Cluster Reload EPAS Sets the maximum time to wait
service for WAL replication.
account
wal_sync_method Cluster Reload EPAS Selects the method used for
service forcing WAL updates to disk.
account
wal_writer_delay Cluster Reload EPAS WAL writer sleep time between
service WAL flushes.
account
work_mem Session Immediate User Sets the maximum memory to
be used for query workspaces.
xloginsert_locks Cluster Restart EPAS Sets the number of locks used
service for concurrent xlog insertions.
account
xmlbinary Session Immediate User Sets how binary values are to
be encoded in XML.
Xmloption Session Immediate User Sets whether XML data in
implicit parsing and
serialization operations is to be
considered as documents or
content fragments.
zero_damaged_pages Session Immediate Superuser Continues processing past
damaged page headers.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 53

 EDB Postgres Advanced Server Guide

3.1.3 Configuration Parameters by Functionality

This section provides more detail for certain groups of configuration parameters.

The section heading for each parameter is followed by a list of attributes:

Parameter Type. Type of values the parameter can accept. See Section 3.1.1 for
a discussion of parameter type values.
Default Value. Default setting if a value is not explicitly set in the configuration
file.
Range. Permitted range of values.
Minimum Scope of Effect. Smallest scope for which a distinct setting can be
made. Generally, the minimal scope of a distinct setting is either the entire cluster
(the setting is the same for all databases and sessions thereof, in the cluster), or
per session (the setting may vary by role, database, or individual session). (This
attribute has the same meaning as the Scope of Effect column in the table of
Section 2.1.2.)
When Value Changes Take Effect. Least invasive action required to activate a
change to a parameters value. All parameter setting changes made in the
configuration file can be put into effect with a restart of the database server;
however certain parameters require a database server restart. Some parameter
setting changes can be put into effect with a reload of the configuration file
without stopping the database server. Finally, other parameter setting changes can
be put into effect with some client side action whose result is immediate. (This
attribute has the same meaning as the When Takes Effect column in the table of
Section 2.1.2.)
Required Authorization to Activate. The type of user authorization to activate a
change to a parameters setting. If a database server restart or a configuration file
reload is required, then the user must be a EPAS service account
(enterprisedb or postgres depending upon the Advanced Server
compatibility installation mode). This attribute has the same meaning as the
Authorized User column in the table of Section 2.1.2.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 54

 EDB Postgres Advanced Server Guide

3.1.3.1 Top Performance Related Parameters

This section discusses the configuration parameters that have the most immediate impact
on performance.

3.1.3.1.1 shared_buffers

Parameter Type: Integer

Default Value: 32MB

Range: 128kB to system dependent

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Restart

Required Authorization to Activate: EPAS service account

Sets the amount of memory the database server uses for shared memory buffers. The
default is typically 32 megabytes (32MB), but might be less if your kernel settings will not
support it (as determined during initdb). This setting must be at least 128 kilobytes.
(Non-default values of BLCKSZ change the minimum.) However, settings significantly
higher than the minimum are usually needed for good performance.

If you have a dedicated database server with 1GB or more of RAM, a reasonable starting
value for shared_buffers is 25% of the memory in your system. There are some
workloads where even large settings for shared_buffers are effective, but because
Advanced Server also relies on the operating system cache, it is unlikely that an
allocation of more than 40% of RAM to shared_buffers will work better than a
smaller amount.

On systems with less than 1GB of RAM, a smaller percentage of RAM is appropriate, so
as to leave adequate space for the operating system (15% of memory is more typical in
these situations). Also, on Windows, large values for shared_buffers aren't as
effective. You may find better results keeping the setting relatively low and using the
operating system cache more instead. The useful range for shared_buffers on
Windows systems is generally from 64MB to 512MB.

Increasing this parameter might cause Advanced Server to request more System V shared
memory than your operating system's default configuration allows. See Section 17.4.1,
Shared Memory and Semaphores in the PostgreSQL Core Documentation for
information on how to adjust those parameters, if necessary.
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 55
 EDB Postgres Advanced Server Guide

3.1.3.1.2 work_mem

Parameter Type: Integer

Default Value: 1MB

Range: 64kB to 2097151kB

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

Specifies the amount of memory to be used by internal sort operations and hash tables
before writing to temporary disk files. The value defaults to one megabyte (1MB). Note
that for a complex query, several sort or hash operations might be running in parallel;
each operation will be allowed to use as much memory as this value specifies before it
starts to write data into temporary files. Also, several running sessions could be doing
such operations concurrently. Therefore, the total memory used could be many times the
value of work_mem; it is necessary to keep this fact in mind when choosing the value.
Sort operations are used for ORDER BY, DISTINCT, and merge joins. Hash tables are
used in hash joins, hash-based aggregation, and hash-based processing of IN subqueries.

Reasonable values are typically between 4MB and 64MB, depending on the size of your
machine, how many concurrent connections you expect (determined by
max_connections), and the complexity of your queries.

3.1.3.1.3 maintenance_work_mem

Parameter Type: Integer

Default Value: 16MB

Range: 1024kB to 2097151kB

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

Specifies the maximum amount of memory to be used by maintenance operations, such

as VACUUM, CREATE INDEX, and ALTER TABLE ADD FOREIGN KEY. It defaults to 16
megabytes (16MB). Since only one of these operations can be executed at a time by a

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 56

 EDB Postgres Advanced Server Guide

database session, and an installation normally doesn't have many of them running
concurrently, it's safe to set this value significantly larger than work_mem. Larger settings
might improve performance for vacuuming and for restoring database dumps.

Note that when autovacuum runs, up to autovacuum_max_workers times this memory

may be allocated, so be careful not to set the default value too high.

A good rule of thumb is to set this to about 5% of system memory, but not more than
about 512MB. Larger values won't necessarily improve performance.

3.1.3.1.4 wal_buffers

Parameter Type: Integer

Default Value: 64kB

Range: 32kB to system dependent

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Restart

Required Authorization to Activate: EPAS service account

The amount of memory used in shared memory for WAL data. The default is 64
kilobytes (64kB). The setting need only be large enough to hold the amount of WAL data
generated by one typical transaction, since the data is written out to disk at every
transaction commit.

Increasing this parameter might cause Advanced Server to request more System V shared
memory than your operating system's default configuration allows. See Section 17.4.1,
Shared Memory and Semaphores in the PostgreSQL Core Documentation for
information on how to adjust those parameters, if necessary.

Although even this very small setting does not always cause a problem, there are
situations where it can result in extra fsync calls, and degrade overall system
throughput. Increasing this value to 1MB or so can alleviate this problem. On very busy
systems, an even higher value may be needed, up to a maximum of about 16MB. Like
shared_buffers, this parameter increases Advanced Servers initial shared memory
allocation, so if increasing it causes an Advanced Server start failure, you will need to
increase the operating system limit.

3.1.3.1.5 checkpoint_segments

Now deprecated; this parameter is not supported by server versions 9.5 or later.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 57

 EDB Postgres Advanced Server Guide

3.1.3.1.6 checkpoint_completion_target

Parameter Type: Floating point

Default Value: 0.5

Range: 0 to 1

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Reload

Required Authorization to Activate: EPAS service account

Specifies the target of checkpoint completion as a fraction of total time between

checkpoints. This spreads out the checkpoint writes while the system starts working
towards the next checkpoint.

The default of 0.5 means aim to finish the checkpoint writes when 50% of the next
checkpoint is ready. A value of 0.9 means aim to finish the checkpoint writes when 90%
of the next checkpoint is done, thus throttling the checkpoint writes over a larger amount
of time and avoiding spikes of performance bottlenecking.

3.1.3.1.7 checkpoint_timeout

Parameter Type: Integer

Default Value: 5min

Range: 30s to 3600s

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Reload

Required Authorization to Activate: EPAS service account

Maximum time between automatic WAL checkpoints, in seconds. The default is five
minutes (5min). Increasing this parameter can increase the amount of time needed for
crash recovery.

Increasing checkpoint_timeout to a larger value, such as 15 minutes, can reduce the

I/O load on your system, especially when using large values for shared_buffers.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 58

 EDB Postgres Advanced Server Guide

The downside of making the aforementioned adjustments to the checkpoint parameters is

that your system will use a modest amount of additional disk space, and will take longer
to recover in the event of a crash. However, for most users, this is a small price to pay for
a significant performance improvement.

3.1.3.1.8 max_wal_size

Parameter Type: Integer

Default Value: 1 GB

Range: 2 to 2147483647

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Reload

Required Authorization to Activate: EPAS service account

max_wal_size specifies the maximum size that the WAL will reach between automatic
WAL checkpoints. This is a soft limit; WAL size can exceed max_wal_size under
special circumstances (when under a heavy load, a failing archive_command, or a high
wal_keep_segments setting).

Increasing this parameter can increase the amount of time needed for crash recovery. This
parameter can only be set in the postgresql.conf file or on the server command line.

3.1.3.1.9 min_wal_size

Parameter Type: Integer

Default Value: 80 MB

Range: 2 to 2147483647

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Reload

Required Authorization to Activate: EPAS service account

If WAL disk usage stays below the value specified by min_wal_size, old WAL files
are recycled for future use at a checkpoint, rather than removed. This ensures that
enough WAL space is reserved to handle spikes in WAL usage (like when running large

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 59

 EDB Postgres Advanced Server Guide

batch jobs). This parameter can only be set in the postgresql.conf file or on the server
command line.

3.1.3.1.10 bgwriter_delay

Parameter Type: Integer

Default Value: 200ms

Range: 10ms to 10000ms

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Reload

Required Authorization to Activate: EPAS service account

Specifies the delay between activity rounds for the background writer. In each round the
writer issues writes for some number of dirty buffers (controllable by the
bgwriter_lru_maxpages and bgwriter_lru_multiplier parameters). It then
sleeps for bgwriter_delay milliseconds, and repeats.

The default value is 200 milliseconds (200ms). Note that on many systems, the effective
resolution of sleep delays is 10 milliseconds; setting bgwriter_delay to a value that is
not a multiple of 10 might have the same results as setting it to the next higher multiple of
10.

Typically, when tuning bgwriter_delay, it should be reduced from its default value.
This parameter is rarely increased, except perhaps to save on power consumption on a
system with very low utilization.

3.1.3.1.11 seq_page_cost

Parameter Type: Floating point

Default Value: 1

Range: 0 to 1.79769e+308

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 60

 EDB Postgres Advanced Server Guide

Sets the planner's estimate of the cost of a disk page fetch that is part of a series of
sequential fetches. The default is 1.0. This value can be overridden for a particular
tablespace by setting the tablespace parameter of the same name. (Refer to the ALTER
TABLESPACE command in the PostgreSQL Core Documentation.)

The default value assumes very little caching, so it's frequently a good idea to reduce it.
Even if your database is significantly larger than physical memory, you might want to try
setting this parameter to less than 1 (rather than its default value of 1) to see whether you
get better query plans that way. If your database fits entirely within memory, you can
lower this value much more, perhaps to 0.1.

3.1.3.1.12 random_page_cost

Parameter Type: Floating point

Default Value: 4

Range: 0 to 1.79769e+308

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

Sets the planner's estimate of the cost of a non-sequentially-fetched disk page. The
default is 4.0. This value can be overridden for a particular tablespace by setting the
tablespace parameter of the same name. (Refer to the ALTER TABLESPACE command in
the PostgreSQL Core Documentation.)

Reducing this value relative to seq_page_cost will cause the system to prefer index
scans; raising it will make index scans look relatively more expensive. You can raise or
lower both values together to change the importance of disk I/O costs relative to CPU
costs, which are described by the cpu_tuple_cost and cpu_index_tuple_cost
parameters.

The default value assumes very little caching, so it's frequently a good idea to reduce it.
Even if your database is significantly larger than physical memory, you might want to try
setting this parameter to 2 (rather than its default of 4) to see whether you get better query
plans that way. If your database fits entirely within memory, you can lower this value
much more, perhaps to 0.1.

Although the system will let you do so, never set random_page_cost less than
seq_page_cost. However, setting them equal (or very close to equal) makes sense if
the database fits mostly or entirely within memory, since in that case there is no penalty

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 61

 EDB Postgres Advanced Server Guide

for touching pages out of sequence. Also, in a heavily-cached database you should lower
both values relative to the CPU parameters, since the cost of fetching a page already in
RAM is much smaller than it would normally be.

3.1.3.1.13 effective_cache_size

Parameter Type: Integer

Default Value: 128MB

Range: 8kB to 17179869176kB

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

Sets the planner's assumption about the effective size of the disk cache that is available to
a single query. This is factored into estimates of the cost of using an index; a higher value
makes it more likely index scans will be used, a lower value makes it more likely
sequential scans will be used. When setting this parameter you should consider both
Advanced Servers shared buffers and the portion of the kernel's disk cache that will be
used for Advanced Server data files. Also, take into account the expected number of
concurrent queries on different tables, since they will have to share the available space.
This parameter has no effect on the size of shared memory allocated by Advanced Server,
nor does it reserve kernel disk cache; it is used only for estimation purposes. The default
is 128 megabytes (128MB).

If this parameter is set too low, the planner may decide not to use an index even when it
would be beneficial to do so. Setting effective_cache_size to 50% of physical
memory is a normal, conservative setting. A more aggressive setting would be
approximately 75% of physical memory.

3.1.3.1.14 synchronous_commit

Parameter Type: Boolean

Default Value: true

Range: {true | false}

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 62

 EDB Postgres Advanced Server Guide

Required Authorization to Activate: Session user

Specifies whether transaction commit will wait for WAL records to be written to disk
before the command returns a "success" indication to the client. The default, and safe,
setting is on. When off, there can be a delay between when success is reported to the
client and when the transaction is really guaranteed to be safe against a server crash. (The
maximum delay is three times wal_writer_delay.)

Unlike fsync, setting this parameter to off does not create any risk of database
inconsistency: an operating system or database crash might result in some recent
allegedly-committed transactions being lost, but the database state will be just the same
as if those transactions had been aborted cleanly.

So, turning synchronous_commit off can be a useful alternative when performance is

more important than exact certainty about the durability of a transaction. See Section
29.3, Asynchronous Commit in the PostgreSQL Core Documentation for information.

This parameter can be changed at any time; the behavior for any one transaction is
determined by the setting in effect when it commits. It is therefore possible, and useful, to
have some transactions commit synchronously and others asynchronously. For example,
to make a single multistatement transaction commit asynchronously when the default is
the opposite, issue SET LOCAL synchronous_commit TO OFF within the
transaction.

3.1.3.1.15 edb_max_spins_per_delay

Parameter Type: Integer

Default Value: 1000

Range: {10 | 1000}

Minimum Scope of Effect: Per cluster

When Value Changes Take Effect: Restart

Required Authorization to Activate: EPAS service account

Use edb_max_spins_per_delay to specify the maximum number of times that a

session will 'spin' while waiting for a spin-lock. If a lock is not acquired, the session will
sleep. If you do not specify an alternative value for edb_max_spins_per_delay, the
server will enforce the default value of 1000.

This may be useful for systems that use NUMA (non-uniform memory access)
architecture.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 63

 EDB Postgres Advanced Server Guide

3.1.3.1.16 pg_prewarm.autoprewarm

Parameter Type: Boolean

Default Value: true

Range: {true | false}

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Restart

Required Authorization to Activate: EPAS service account

This parameter controls whether or not the database server should run autoprewarm,
which is a background worker process that automatically dumps shared buffers to disk
before a shutdown. It then prewarms the shared buffers the next time the server is started,
meaning it loads blocks from the disk back into the buffer pool.

The advantage is that it shortens the warm up times after the server has been restarted by
loading the data that has been dumped to disk before shutdown.

If pg_prewarm.autoprewarm is set to on, the autoprewarm worker is enabled. If the

parameter is set to off, autoprewarm is disabled. The parameter is on by default.

Before autoprewarm can be used, you must add $libdir/pg_prewarm to the

libraries listed in the shared_preload_libraries configuration parameter of the
postgresql.conf file as shown by the following example:

shared_preload_libraries =
'$libdir/dbms_pipe,$libdir/edb_gen,$libdir/dbms_aq,$libdir/pg_prewarm'

After modifying the shared_preload_libraries parameter, restart the database

server after which the autoprewarm background worker is launched immediately after
the server has reached a consistent state.

The autoprewarm process will start loading blocks that were previously recorded in
$PGDATA/autoprewarm.blocks until there is no free buffer space left in the buffer
pool. In this manner, any new blocks that were loaded either by the recovery process or
by the querying clients, are not replaced.

Once the autoprewarm process has finished loading buffers from disk, it will
periodically dump shared buffers to disk at the interval specified by the
pg_prewarm.autoprewarm_interval parameter (see Section 3.1.3.1.17). Upon the
next server restart, the autoprewarm process will prewarm shared buffers with the
blocks that were last dumped to disk.
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 64
 EDB Postgres Advanced Server Guide

3.1.3.1.17 pg_prewarm.autoprewarm_interval

Parameter Type: Integer

Default Value: 300s

Range: 0s to 2147483s

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Reload

Required Authorization to Activate: EPAS service account

This is the minimum number of seconds after which the autoprewarm background
worker dumps shared buffers to disk. The default is 300 seconds. If set to 0, shared
buffers are not dumped at regular intervals, but only when the server is shut down.

See Section 3.1.3.1.16 for information on the autoprewarm background worker.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 65

 EDB Postgres Advanced Server Guide

3.1.3.2 Resource Usage / Memory

The configuration parameters in this section control resource usage pertaining to
memory.

3.1.3.2.1 edb_dynatune

Parameter Type: Integer

Default Value: 0

Range: 0 to 100

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Restart

Required Authorization to Activate: EPAS service account

Determines how much of the host systems resources are to be used by the database
server based upon the host machines total available resources and the intended usage of
the host machine.

When Advanced Server is initially installed, the edb_dynatune parameter is set in

accordance with the selected usage of the host machine on which it was installed (i.e.,
development machine, mixed use machine, or dedicated server). For most purposes, there
is no need for the database administrator to adjust the various configuration parameters in
the postgresql.conf file in order to improve performance.

The edb_dynatune parameter can be set to any integer value between 0 and 100,
inclusive. A value of 0, turns off the dynamic tuning feature thereby leaving the database
server resource usage totally under the control of the other configuration parameters in
the postgresql.conf file.

A low non-zero, value (e.g., 1 - 33) dedicates the least amount of the host machines
resources to the database server. This setting would be used for a development machine
where many other applications are being used.

A value in the range of 34 - 66 dedicates a moderate amount of resources to the database

server. This setting might be used for a dedicated application server that may have a fixed
number of other applications running on the same machine as Advanced Server.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 66

 EDB Postgres Advanced Server Guide

The highest values (e.g., 67 - 100) dedicate most of the servers resources to the database
server. This setting would be used for a host machine that is totally dedicated to running
Advanced Server.

Once a value of edb_dynatune is selected, database server performance can be further

fine-tuned by adjusting the other configuration parameters in the postgresql.conf
file. Any adjusted setting overrides the corresponding value chosen by edb_dynatune.
You can change the value of a parameter by un-commenting the configuration parameter,
specifying the desired value, and restarting the database server.

3.1.3.2.2 edb_dynatune_profile

Parameter Type: Enum

Default Value: oltp

Range: {oltp | reporting | mixed}

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Restart

Required Authorization to Activate: EPAS service account

This parameter is used to control tuning aspects based upon the expected workload
profile on the database server.

The following are the possible values:

oltp. Recommended when the database server is processing heavy online

transaction processing workloads.
reporting. Recommended for database servers used for heavy data reporting.
mixed. Recommended for servers that provide a mix of transaction processing
and data reporting.

3.1.3.2.3 edb_enable_icache

Parameter Type: Boolean

Default Value: false

Range: {true | false}

Minimum Scope of Effect: Cluster

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 67

 EDB Postgres Advanced Server Guide

When Value Changes Take Effect: Restart

Required Authorization to Activate: EPAS service account

Enables or disables Infinite Cache. If edb_enable_icache is set to on, Infinite Cache

is enabled; if the parameter is set to off, Infinite Cache is disabled.

If you set edb_enable_icache to on, you must also specify a list of cache servers by
setting the edb_icache_servers parameter.

3.1.3.2.4 edb_icache_servers

Parameter Type: String

Default Value: none

Range: n/a

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Reload

Required Authorization to Activate: EPAS service account

The edb_icache_servers parameter specifies a list of one or more servers with active
edb-icache daemons. edb_icache_servers is a string value that takes the form of a
comma-separated list of hostname:port pairs. You can specify each pair in any of the
following forms:

hostname
IP_address
hostname:portnumber
IP_address:portnumber

If you do not specify a port number, Infinite Cache assumes that the cache server is
listening at port 11211. This configuration parameter will take effect only if
edb_enable_icache is set to on. Use the edb_icache_servers parameter to specify
a maximum of 128 cache nodes.

You can dynamically modify the Infinite Cache server nodes. To change the Infinite
Cache server configuration, use the edb_icache_servers parameter in the
postgresql.conf file to perform the following:

Specify additional cache information to add server(s).

Delete server information to remove server(s).
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 68
 EDB Postgres Advanced Server Guide

Specify additional server information and delete existing server information to

both add and delete servers during the same reload operation.

After updating the edb_icache_servers parameter in the postgresql.conf file,

you must reload the configuration parameters for the changes to take effect.

3.1.3.2.5 edb_icache_compression_level

Parameter Type: Integer

Default Value: 6

Range: 0 to 9

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Superuser

The edb_icache_compression_level parameter controls the compression level that

is applied to each page before storing it in the distributed Infinite Cache.

When Advanced Server reads data from disk, it typically reads the data in 8kB
increments. If edb_icache_compression_level is set to 0, each time Advanced
Server sends an 8kB page to the Infinite Cache server that page is stored (uncompressed)
in 8kB of cache memory. If the edb_icache_compression_level parameter is set to
9, Advanced Server applies the maximum compression possible before sending it to the
Infinite Cache server, so a page that previously took 8kB of cached memory might take
2kB of cached memory. Exact compression numbers are difficult to predict, as they are
dependent on the nature of the data on each page.

This parameter must be an integer in the range 0 to 9.

A compression level of 0 disables compression; it uses no CPU time for

compression, but requires more storage space and network resources to process.
A compression level of 9 invokes the maximum amount of compression; it
increases the load on the CPU, but less data flows across the network, so network
demand is reduced. Each page takes less room on the Infinite Cache server, so
memory requirements are reduced.
A compression level of 5 or 6 is a reasonable compromise between the amount of
compression received and the amount of CPU time invested.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 69

 EDB Postgres Advanced Server Guide

The compression level must be set by the superuser and can be changed for the current
session while the server is running. The following command disables the compression
mechanism for the currently active session:

SET edb_icache_compression_level TO 0;

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 70

 EDB Postgres Advanced Server Guide

3.1.3.3 Resource Usage / EDB Resource Manager

The configuration parameters in this section control resource usage through EDB
Resource Manager.

3.1.3.3.1 edb_max_resource_groups

Parameter Type: Integer

Default Value: 16

Range: 0 to 65536

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Restart

Required Authorization to Activate: EPAS service account

This parameter controls the maximum number of resource groups that can be used
simultaneously by EDB Resource Manager. More resource groups can be created than
the value specified by edb_max_resource_groups, however, the number of resource
groups in active use by processes in these groups cannot exceed this value.

Parameter edb_max_resource_groups should be set comfortably larger than the

number of groups you expect to maintain so as not to run out.

3.1.3.3.2 edb_resource_group

Parameter Type: String

Default Value: none

Range: n/a

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 71

 EDB Postgres Advanced Server Guide

Set the edb_resource_group parameter to the name of the resource group to which
the current session is to be controlled by EDB Resource Manager according to the
groups resource type settings.

If the parameter is not set, then the current session does not utilize EDB Resource
Manager.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 72

 EDB Postgres Advanced Server Guide

3.1.3.4 Query Tuning

This section describes the configuration parameters used for optimizer hints.

3.1.3.4.1 enable_hints

Parameter Type: Boolean

Default Value: true

Range: {true | false}

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

Optimizer hints embedded in SQL commands are utilized when enable_hints is on.
Optimizer hints are ignored when this parameter is off.

3.1.3.5 Query Tuning / Planner Method Configuration

This section describes the configuration parameters used for planner method
configuration.

3.1.3.5.1 edb_custom_plan_tries

Parameter Type: Numeric

Default Value: 5

Range: {0 | 100}

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session User

This configuration parameter controls the number of custom execution plans considered
by the planner before the planner settles on a generic execution plan.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 73

 EDB Postgres Advanced Server Guide

When a client application repeatedly executes a prepared statement, the server may
decide to evaluate several execution plans before deciding to choose a custom plan or a
generic plan.
A custom plan is a plan built for a specific set of parameter values.
A generic plan is a plan that will work with any set of parameter values supplied
by the client application.

By default, the optimizer will generate five custom plans before evaluating a generic
plan. That means that if you execute a prepared statement six times, the optimizer will
generate five custom plans, then one generic plan, and then decide whether to stick with
the generic plan.

In certain workloads, this extra planning can have a negative impact on performance.
You can adjust the edb_custom_plan_tries configuration parameter to decrease the
number of custom plans considered before evaluating a generic plan. Setting
edb_custom_plan_tries to 0 will effectively disable custom plan generation.

Consider the following query:

PREPARE custQuery AS SELECT * FROM customer WHERE salesman >= $1

The $1 token in this query is a parameter marker - the client application must provide a
value for each parameter marker each time the statement executes.

If an index has been defined on customer.salesman, the optimizer may choose to

execute this query using a sequential scan, or using an index scan. In some cases, an
index is faster than a sequential scan; in other cases, the sequential scan will win. The
optimal plan will depend on the distribution of salesman values in the table and on the
search value (the value provided for the $1 parameter).

When the client application repeatedly executes the custQuery prepared statement, the
optimizer will generate some number of parameter-value-specific execution plans
(custom plans), followed by a generic plan (a plan that ignores the parameter values), and
then decide whether to stick with the generic plan or to continue to generate custom plans
for each execution. The decision process takes into account not only the cost of executing
the plans, but the cost of generating custom plans as well.

3.1.3.5.2 edb_enable_pruning

Parameter Type: Boolean

Default Value: true

Range: {true | false}

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 74

 EDB Postgres Advanced Server Guide

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

When set to TRUE, edb_enable_pruning allows the query planner to early-prune

partitioned tables. Early-pruning means that the query planner can prune (i.e., ignore)
partitions that would not be searched in a query before generating query plans. This helps
improve performance time as it eliminates the generation of query plans of partitions that
would not be searched.

Conversely, late-pruning means that the query planner prunes partitions after generating
query plans for each partition. (The constraint_exclusion configuration parameter
controls late-pruning.)

The ability to early-prune depends upon the nature of the query in the WHERE clause.
Early-pruning can be utilized in only simple queries with constraints of the type WHERE
column = literal (e.g., WHERE deptno = 10).

Early-pruning is not used for more complex queries such as WHERE column =
expression (e.g., WHERE deptno = 10 + 5).

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 75

 EDB Postgres Advanced Server Guide

3.1.3.6 Reporting and Logging / What to Log

The configuration parameters in this section control reporting and logging.

3.1.3.6.1 trace_hints

Parameter Type: Boolean

Default Value: false

Range: {true | false}

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

Use with the optimizer hints feature to provide more detailed information regarding
whether or not a hint was used by the planner. Set the client_min_messages and
trace_hints configuration parameters as follows:

SET client_min_messages TO info;

SET trace_hints TO true;

The SELECT command with the NO_INDEX hint shown below illustrates the additional
information produced when the aforementioned configuration parameters are set.

EXPLAIN SELECT /*+ NO_INDEX(accounts accounts_pkey) */ * FROM accounts WHERE

aid = 100;

INFO: [HINTS] Index Scan of [accounts].[accounts_pkey] rejected because of

NO_INDEX hint.

INFO: [HINTS] Bitmap Heap Scan of [accounts].[accounts_pkey] rejected

because of NO_INDEX hint.
QUERY PLAN
-------------------------------------------------------------
Seq Scan on accounts (cost=0.00..14461.10 rows=1 width=97)
Filter: (aid = 100)
(2 rows)

3.1.3.6.2 edb_log_every_bulk_value

Parameter Type: Boolean

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 76

 EDB Postgres Advanced Server Guide

Default Value: false

Range: {true | false}

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Superuser

Bulk processing logs the resulting statements into both the Advanced Server log file and
the EDB Audit log file. However, logging each and every statement in bulk processing is
costly. This can be controlled by the edb_log_every_bulk_value configuration
parameter. When set to true, each and every statement in bulk processing is logged.
When set to false, a log message is recorded once per bulk processing. In addition, the
duration is emitted once per bulk processing. Default is set to false.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 77

 EDB Postgres Advanced Server Guide

3.1.3.7 Auditing Settings

This section describes configuration parameters used by the Advanced Server database
auditing feature.

3.1.3.7.1 edb_audit

Parameter Type: Enum

Default Value: none

Range: {none | csv | xml}

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Reload

Required Authorization to Activate: EPAS service account

Enables or disables database auditing. The values xml or csv will enable database
auditing. These values represent the file format in which auditing information will be
captured. none will disable database auditing and is also the default.

3.1.3.7.2 edb_audit_directory

Parameter Type: String

Default Value: edb_audit

Range: n/a

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Reload

Required Authorization to Activate: EPAS service account

Specifies the directory where the audit log files will be created. The path of the directory
can be absolute or relative to the Advanced Server data directory.

3.1.3.7.3 edb_audit_filename

Parameter Type: String

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 78
 EDB Postgres Advanced Server Guide

Default Value: audit-%Y%m%d_%H%M%S

Range: n/a

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Reload

Required Authorization to Activate: EPAS service account

Specifies the file name of the audit file where the auditing information will be stored. The
default file name will be audit-%Y%m%d_%H%M%S. The escape sequences, %Y, %m etc.,
will be replaced by the appropriate current values according to the system date and time.

3.1.3.7.4 edb_audit_rotation_day

Parameter Type: String

Default Value: every

Range: {none | every | sun | mon | tue | wed | thu | fri | sat} ...

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Reload

Required Authorization to Activate: EPAS service account

Specifies the day of the week on which to rotate the audit files. Valid values are sun,
mon, tue, wed, thu, fri, sat, every, and none. To disable rotation, set the value to
none. To rotate the file every day, set the edb_audit_rotation_day value to every.
To rotate the file on a specific day of the week, set the value to the desired day of the
week.

3.1.3.7.5 edb_audit_rotation_size

Parameter Type: Integer

Default Value: 0MB

Range: 0MB to 5000MB

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Reload

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 79
 EDB Postgres Advanced Server Guide

Required Authorization to Activate: EPAS service account

Specifies a file size threshold in megabytes when file rotation will be forced to occur. The
default value is 0MB. If the parameter is commented out or set to 0, rotation of the file on
a size basis will not occur.

3.1.3.7.6 edb_audit_rotation_seconds

Parameter Type: Integer

Default Value: 0s

Range: 0s to 2147483647s

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Reload

Required Authorization to Activate: EPAS service account

Specifies the rotation time in seconds when a new log file should be created. To disable
this feature, set this parameter to 0.

3.1.3.7.7 edb_audit_connect

Parameter Type: Enum

Default Value: failed

Range: {none | failed | all}

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Reload

Required Authorization to Activate: EPAS service account

Enables auditing of database connection attempts by users. To disable auditing of all

connection attempts, set edb_audit_connect to none. To audit all failed connection
attempts, set the value to failed. To audit all connection attempts, set the value to all.

3.1.3.7.8 edb_audit_disconnect

Parameter Type: Enum

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 80

 EDB Postgres Advanced Server Guide

Default Value: none

Range: {none | all}

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Reload

Required Authorization to Activate: EPAS service account

Enables auditing of database disconnections by connected users. To enable auditing of

disconnections, set the value to all. To disable, set the value to none.

3.1.3.7.9 edb_audit_statement

Parameter Type: String

Default Value: ddl, error

Range: {none | ddl | dml | insert | update | delete | truncate | select | error |
create | drop | alter | grant | revoke | rollback | all} ...

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Reload

Required Authorization to Activate: EPAS service account

This configuration parameter is used to specify auditing of different categories of SQL

statements as well as those statements related to specific SQL commands. To log errors,
set the parameter value to error. To audit all DDL statements such as CREATE TABLE,
ALTER TABLE, etc., set the parameter value to ddl. To audit specific types of DDL
statements, the parameter values can include those specific SQL commands (create,
drop, or alter). In addition, the object type may be specified following the command
such as create table, create view, drop role, etc. All modification statements
such as INSERT, UPDATE, DELETE or TRUNCATE can be audited by setting
edb_audit_statement to dml. To audit specific types of DML statements, the
parameter values can include the specific SQL commands, insert, update, delete, or
truncate. Include parameter values select, grant, revoke, or rollback to audit
statements regarding those SQL commands. Setting the value to all will audit every
statement while none disables this feature.

3.1.3.7.10 edb_audit_tag

Parameter Type: String

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 81
 EDB Postgres Advanced Server Guide

Default Value: none

Minimum Scope of Effect: Session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: User

Use edb_audit_tag to specify a string value that will be included in the audit log when
the edb_audit parameter is set to csv or xml.

3.1.3.7.11 edb_audit_destination

Parameter Type: Enum

Default Value: file

Range: {file | syslog}

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Reload

Required Authorization to Activate: EPAS service account

Specifies whether the audit log information is to be recorded in the directory as given by
the edb_audit_directory parameter or to the directory and file managed by the
syslog process. Set to file to use the directory specified by edb_audit_directory
(the default setting). Set to syslog to use the syslog process and its location as
configured in the /etc/syslog.conf file. Note: In recent Linux versions, syslog has
been replaced by rsyslog and the configuration file is in /etc/rsyslog.conf.

3.1.3.7.12 edb_log_every_bulk_value

For information on edb_log_every_bulk_value, see Section 3.1.3.6.2.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 82

 EDB Postgres Advanced Server Guide

3.1.3.8 Client Connection Defaults / Locale and Formatting

This section describes configuration parameters affecting locale and formatting.

3.1.3.8.1 icu_short_form

Parameter Type: String

Default Value: none

Range: n/a

Minimum Scope of Effect: Database

When Value Changes Take Effect: n/a

Required Authorization to Activate: n/a

The configuration parameter icu_short_form is a parameter containing the default

ICU short form name assigned to a database or to the Advanced Server instance. See
Section 3.6 for general information about the ICU short form and the Unicode Collation
Algorithm.

This configuration parameter is set either when the CREATE DATABASE command is
used with the ICU_SHORT_FORM parameter (see Section 3.6.3.2) in which case the
specified short form name is set and appears in the icu_short_form configuration
parameter when connected to this database, or when an Advanced Server instance is
created with the initdb command used with the --icu_short_form option (see
Section 3.6.3.3) in which case the specified short form name is set and appears in the
icu_short_form configuration parameter when connected to a database in that
Advanced Server instance, and the database does not override it with its own
ICU_SHORT_FORM parameter with a different short form.

Once established in the manner described, the icu_short_form configuration

parameter cannot be changed.

3.1.3.9 Client Connection Defaults / Statement Behavior

This section describes configuration parameters affecting statement behavior.

3.1.3.9.1 default_heap_fillfactor

Parameter Type: Integer

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 83
 EDB Postgres Advanced Server Guide

Default Value: 100

Range: 10 to 100

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

Sets the fillfactor for a table when the FILLFACTOR storage parameter is omitted from a
CREATE TABLE command.

The fillfactor for a table is a percentage between 10 and 100. 100 (complete packing) is
the default. When a smaller fillfactor is specified, INSERT operations pack table pages
only to the indicated percentage; the remaining space on each page is reserved for
updating rows on that page. This gives UPDATE a chance to place the updated copy of a
row on the same page as the original, which is more efficient than placing it on a different
page. For a table whose entries are never updated, complete packing is the best choice,
but in heavily updated tables smaller fillfactors are appropriate.

3.1.3.10 Client Connection Defaults / Other Defaults

The parameters in this section set other miscellaneous client connection defaults.

3.1.3.10.1 oracle_home

Parameter Type: String

Default Value: none

Range: n/a

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Restart

Required Authorization to Activate: EPAS service account

Before creating an Oracle Call Interface (OCI) database link to an Oracle server, you
must direct Advanced Server to the correct Oracle home directory. Set the
LD_LIBRARY_PATH environment variable on Linux (or PATH on Windows) to the lib
directory of the Oracle client installation directory.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 84

 EDB Postgres Advanced Server Guide

For Windows only, you can instead set the value of the oracle_home configuration
parameter in the postgresql.conf file. The value specified in the oracle_home
configuration parameter will override the Windows PATH environment variable.

The LD_LIBRARY_PATH environment variable on Linux (PATH environment variable or

oracle_home configuration parameter on Windows) must be set properly each time you
start Advanced Server.

For Windows only: To set the oracle_home configuration parameter in the

postgresql.conf file, edit the file, adding the following line:

oracle_home = 'lib_directory'

Substitute the name of the Windows directory that contains oci.dll for
lib_directory.

After setting the oracle_home configuration parameter, you must restart the server for
the changes to take effect. Restart the server from the Windows Services console.

3.1.3.10.2 odbc_lib_path

Parameter Type: String

Default Value: none

Range: n/a

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Restart

Required Authorization to Activate: EPAS service account

If you will be using an ODBC driver manager, and if it is installed in a non-standard

location, you specify the location by setting the odbc_lib_path configuration
parameter in the postgresql.conf file:

odbc_lib_path = 'complete_path_to_libodbc.so'

The configuration file must include the complete pathname to the driver manager shared
library (typically libodbc.so).

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 85

 EDB Postgres Advanced Server Guide

3.1.3.11 Compatibility Options

The configuration parameters described in this section control various database
compatibility features.

3.1.3.11.1 edb_redwood_date

Parameter Type: Boolean

Default Value: false

Range: {true | false}

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

When DATE appears as the data type of a column in the commands, it is translated to
TIMESTAMP(0) at the time the table definition is stored in the database if the
configuration parameter edb_redwood_date is set to TRUE. Thus, a time component
will also be stored in the column along with the date.

If edb_redwood_date is set to FALSE the columns data type in a CREATE TABLE or

ALTER TABLE command remains as a native PostgreSQL DATE data type and is stored as
such in the database. The PostgreSQL DATE data type stores only the date without a time
component in the column.

Regardless of the setting of edb_redwood_date, when DATE appears as a data type in

any other context such as the data type of a variable in an SPL declaration section, or the
data type of a formal parameter in an SPL procedure or SPL function, or the return type
of an SPL function, it is always internally translated to a TIMESTAMP(0) and thus, can
handle a time component if present.

3.1.3.11.2 edb_redwood_greatest_least

Parameter Type: Boolean

Default Value: true

Range: {true | false}

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 86

 EDB Postgres Advanced Server Guide

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

The GREATEST function returns the parameter with the greatest value from its list of
parameters. The LEAST function returns the parameter with the least value from its list of
parameters.

When edb_redwood_greatest_least is set to TRUE, the GREATEST and LEAST

functions return null when at least one of the parameters is null.

SET edb_redwood_greatest_least TO on;

SELECT GREATEST(1, 2, NULL, 3);

greatest
----------

(1 row)

When edb_redwood_greatest_least is set to FALSE, null parameters are ignored

except when all parameters are null in which case null is returned by the functions.

SET edb_redwood_greatest_least TO off;

SELECT GREATEST(1, 2, NULL, 3);

greatest
----------
3
(1 row)

SELECT GREATEST(NULL, NULL, NULL);

greatest
----------

(1 row)

3.1.3.11.3 edb_redwood_raw_names

Parameter Type: Boolean

Default Value: false

Range: {true | false}

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 87
 EDB Postgres Advanced Server Guide

Required Authorization to Activate: Session user

When edb_redwood_raw_names is set to its default value of FALSE, database object

names such as table names, column names, trigger names, program names, user names,
etc. appear in uppercase letters when viewed from Redwood catalogs (that is, system
catalogs prefixed by ALL_, DBA_, or USER_). In addition, quotation marks enclose names
that were created with enclosing quotation marks.

When edb_redwood_raw_names is set to TRUE, the database object names are

displayed exactly as they are stored in the PostgreSQL system catalogs when viewed
from the Redwood catalogs. Thus, names created without enclosing quotation marks
appear in lowercase as expected in PostgreSQL. Names created with enclosing quotation
marks appear exactly as they were created, but without the quotation marks.

For example, the following user name is created, and then a session is started with that
user.

CREATE USER reduser IDENTIFIED BY password;

edb=# \c - reduser
Password for user reduser:
You are now connected to database "edb" as user "reduser".

When connected to the database as reduser, the following tables are created.

CREATE TABLE all_lower (col INTEGER);

CREATE TABLE ALL_UPPER (COL INTEGER);
CREATE TABLE "Mixed_Case" ("Col" INTEGER);

When viewed from the Redwood catalog, USER_TABLES, with

edb_redwood_raw_names set to the default value FALSE, the names appear in
uppercase except for the Mixed_Case name, which appears as created and also with
enclosing quotation marks.

edb=> SELECT * FROM USER_TABLES;

schema_name | table_name | tablespace_name | status | temporary
-------------+--------------+-----------------+--------+-----------
REDUSER | ALL_LOWER | | VALID | N
REDUSER | ALL_UPPER | | VALID | N
REDUSER | "Mixed_Case" | | VALID | N
(3 rows)

When viewed with edb_redwood_raw_names set to TRUE, the names appear in

lowercase except for the Mixed_Case name, which appears as created, but now without
the enclosing quotation marks.

edb=> SET edb_redwood_raw_names TO true;

SET
edb=> SELECT * FROM USER_TABLES;
schema_name | table_name | tablespace_name | status | temporary
-------------+------------+-----------------+--------+-----------
reduser | all_lower | | VALID | N
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 88
 EDB Postgres Advanced Server Guide

reduser | all_upper | | VALID | N

reduser | Mixed_Case | | VALID | N
(3 rows)

These names now match the case when viewed from the PostgreSQL pg_tables
catalog.

edb=> SELECT schemaname, tablename, tableowner FROM pg_tables WHERE

tableowner = 'reduser';
schemaname | tablename | tableowner
------------+------------+------------
reduser | all_lower | reduser
reduser | all_upper | reduser
reduser | Mixed_Case | reduser
(3 rows)

3.1.3.11.4 edb_redwood_strings

Parameter Type: Boolean

Default Value: false

Range: {true | false}

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

If the edb_redwood_strings parameter is set to TRUE, when a string is concatenated

with a null variable or null column, the result is the original string. If
edb_redwood_strings is set to FALSE, the native PostgreSQL behavior is maintained,
which is the concatenation of a string with a null variable or null column gives a null
result.

The following example illustrates the difference.

The sample application contains a table of employees. This table has a column named
comm that is null for most employees. The following query is run with
edb_redwood_string set to FALSE. The concatenation of a null column with non-
empty strings produces a final result of null, so only employees that have a commission
appear in the query result. The output line for all other employees is null.

SET edb_redwood_strings TO off;

SELECT RPAD(ename,10) || ' ' || TO_CHAR(sal,'99,999.99') || ' ' ||

TO_CHAR(comm,'99,999.99') "EMPLOYEE COMPENSATION" FROM emp;

EMPLOYEE COMPENSATION

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 89

 EDB Postgres Advanced Server Guide

----------------------------------

ALLEN 1,600.00 300.00

WARD 1,250.00 500.00

MARTIN 1,250.00 1,400.00

TURNER 1,500.00 .00

(14 rows)

The following is the same query executed when edb_redwood_strings is set to TRUE.
Here, the value of a null column is treated as an empty string. The concatenation of an
empty string with a non-empty string produces the non-empty string.

SET edb_redwood_strings TO on;

SELECT RPAD(ename,10) || ' ' || TO_CHAR(sal,'99,999.99') || ' ' ||

TO_CHAR(comm,'99,999.99') "EMPLOYEE COMPENSATION" FROM emp;

EMPLOYEE COMPENSATION
----------------------------------
SMITH 800.00
ALLEN 1,600.00 300.00
WARD 1,250.00 500.00
JONES 2,975.00
MARTIN 1,250.00 1,400.00
BLAKE 2,850.00
CLARK 2,450.00
SCOTT 3,000.00
KING 5,000.00
TURNER 1,500.00 .00
ADAMS 1,100.00
JAMES 950.00
FORD 3,000.00
MILLER 1,300.00
(14 rows)

3.1.3.11.5 edb_stmt_level_tx

Parameter Type: Boolean

Default Value: false

Range: {true | false}

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 90
 EDB Postgres Advanced Server Guide

The term statement level transaction isolation describes the behavior whereby when a
runtime error occurs in a SQL command, all the updates on the database caused by that
single command are rolled back. For example, if a single UPDATE command successfully
updates five rows, but an attempt to update a sixth row results in an exception, the
updates to all six rows made by this UPDATE command are rolled back. The effects of
prior SQL commands that have not yet been committed or rolled back are pending until a
COMMIT or ROLLBACK command is executed.

In Advanced Server, if an exception occurs while executing a SQL command, all the
updates on the database since the start of the transaction are rolled back. In addition, the
transaction is left in an aborted state and either a COMMIT or ROLLBACK command must
be issued before another transaction can be started.

If edb_stmt_level_tx is set to TRUE, then an exception will not automatically roll

back prior uncommitted database updates. If edb_stmt_level_tx is set to FALSE, then
an exception will roll back uncommitted database updates.

Note: Use edb_stmt_level_tx set to TRUE only when absolutely necessary, as this
may cause a negative performance impact.

The following example run in PSQL shows that when edb_stmt_level_tx is FALSE,
the abort of the second INSERT command also rolls back the first INSERT command.
Note that in PSQL, the command \set AUTOCOMMIT off must be issued, otherwise
every statement commits automatically defeating the purpose of this demonstration of the
effect of edb_stmt_level_tx.

\set AUTOCOMMIT off

SET edb_stmt_level_tx TO off;

INSERT INTO emp (empno,ename,deptno) VALUES (9001, 'JONES', 40);

INSERT INTO emp (empno,ename,deptno) VALUES (9002, 'JONES', 00);
ERROR: insert or update on table "emp" violates foreign key constraint
"emp_ref_dept_fk"
DETAIL: Key (deptno)=(0) is not present in table "dept".

COMMIT;
SELECT empno, ename, deptno FROM emp WHERE empno > 9000;

empno | ename | deptno

-------+-------+--------
(0 rows)

In the following example, with edb_stmt_level_tx set to TRUE, the first INSERT
command has not been rolled back after the error on the second INSERT command. At
this point, the first INSERT command can either be committed or rolled back.

\set AUTOCOMMIT off

SET edb_stmt_level_tx TO on;

INSERT INTO emp (empno,ename,deptno) VALUES (9001, 'JONES', 40);

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 91

 EDB Postgres Advanced Server Guide

INSERT INTO emp (empno,ename,deptno) VALUES (9002, 'JONES', 00);

ERROR: insert or update on table "emp" violates foreign key constraint
"emp_ref_dept_fk"
DETAIL: Key (deptno)=(0) is not present in table "dept"

SELECT empno, ename, deptno FROM emp WHERE empno > 9000;

empno | ename | deptno

-------+-------+--------
9001 | JONES | 40
(1 row)

COMMIT;

A ROLLBACK command could have been issued instead of the COMMIT command in
which case the insert of employee number 9001 would have been rolled back as well.

3.1.3.11.6 db_dialect

Parameter Type: Enum

Default Value: postgres

Range: {postgres | redwood}

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

In addition to the native PostgreSQL system catalog, pg_catalog, Advanced Server

contains extended catalog views as well as system catalogs compatible with Microsoft
SQL Server. These are sys for the expanded catalog views and dbo for SQL Server.
The db_dialect parameter controls the order in which these catalogs are searched for
name resolution.

When set to postgres, the namespace precedence is pg_catalog, sys, then dbo,
giving the PostgreSQL catalog the highest precedence. When set to redwood, the
namespace precedence is sys, dbo, then pg_catalog, giving the expanded catalog
views the highest precedence.

3.1.3.11.7 default_with_rowids

Parameter Type: Boolean

Default Value: false

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 92

 EDB Postgres Advanced Server Guide

Range: {true | false}

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

When set to on, CREATE TABLE includes a ROWID column in newly created tables,
which can then be referenced in SQL commands.

3.1.3.11.8 optimizer_mode

Parameter Type: Enum

Default Value: choose

Range: {choose | ALL_ROWS | FIRST_ROWS | FIRST_ROWS_10 | FIRST_ROWS_100 |

FIRST_ROWS_1000}

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

Sets the default optimization mode for analyzing optimizer hints.

The following table shows the possible values:

Table 3-2 - Optimizer Modes

Hint Description
ALL_ROWS Optimizes for retrieval of all rows of the result set.
Does no default optimization based on assumed number of rows to be retrieved
CHOOSE
from the result set. This is the default.
FIRST_ROWS Optimizes for retrieval of only the first row of the result set.
FIRST_ROWS_10 Optimizes for retrieval of the first 10 rows of the results set.
FIRST_ROWS_100 Optimizes for retrieval of the first 100 rows of the result set.
FIRST_ROWS_1000 Optimizes for retrieval of the first 1000 rows of the result set.

These optimization modes are based upon the assumption that the client submitting the
SQL command is interested in viewing only the first n rows of the result set and will
then abandon the remainder of the result set. Resources allocated to the query are
adjusted as such.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 93

 EDB Postgres Advanced Server Guide

3.1.3.12 Customized Options

In previous releases of Advanced Server, the custom_variable_classes was

required by those parameters not normally known to be added by add-on modules (such
as procedural languages).

3.1.3.12.1 custom_variable_classes

The custom_variable_classes parameter is deprecated in Advanced Server 9.2;

parameters that previously depended on this parameter no longer require its support.

3.1.3.12.2 dbms_alert.max_alerts

Parameter Type: Integer

Default Value: 100

Range: 0 to 500

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Restart

Required Authorization to Activate: EPAS service account

Specifies the maximum number of concurrent alerts allowed on a system using the
DBMS_ALERTS package.

3.1.3.12.3 dbms_pipe.total_message_buffer

Parameter Type: Integer

Default Value: 30 Kb

Range: 30 Kb to 256 Kb

Minimum Scope of Effect: Postmaster

When Value Changes Take Effect: Restart

Required Authorization to Activate: EPAS service account

Specifies the total size of the buffer used for the DBMS_PIPE package.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 94

 EDB Postgres Advanced Server Guide

3.1.3.12.4 index_advisor.enabled

Parameter Type: Boolean

Default Value: true

Range: {true | false}

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

Provides the capability to temporarily suspend Index Advisor in an EDB-PSQL or PSQL

session. The Index Advisor plugin, index_advisor, must be loaded in the EDB-PSQL
or PSQL session in order to use the index_advisor.enabled configuration
parameter.

The Index Advisor plugin can be loaded as shown by the following example:

$ psql -d edb -U enterprisedb

Password for user enterprisedb:
psql (10.0.1)
Type "help" for help.

edb=# LOAD 'index_advisor';

LOAD

Use the SET command to change the parameter setting to control whether or not Index
Advisor generates an alternative query plan as shown by the following example:

edb=# SET index_advisor.enabled TO off;

SET
edb=# EXPLAIN SELECT * FROM t WHERE a < 10000;
QUERY PLAN
-------------------------------------------------------
Seq Scan on t (cost=0.00..1693.00 rows=9864 width=8)
Filter: (a < 10000)
(2 rows)

edb=# SET index_advisor.enabled TO on;

SET
edb=# EXPLAIN SELECT * FROM t WHERE a < 10000;
QUERY PLAN
-----------------------------------------------------------------------------
Seq Scan on t (cost=0.00..1693.00 rows=9864 width=8)
Filter: (a < 10000)
Result (cost=0.00..327.88 rows=9864 width=8)
One-Time Filter: '===[ HYPOTHETICAL PLAN ]==='::text
-> Index Scan using "<hypothetical-index>:1" on t (cost=0.00..327.88
rows=9864 width=8)
Index Cond: (a < 10000)
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 95
 EDB Postgres Advanced Server Guide

(6 rows)

3.1.3.12.5 edb_sql_protect.enabled

Parameter Type: Boolean

Default Value: false

Range: {true | false}

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Reload

Required Authorization to Activate: EPAS service account

Controls whether or not SQL/Protect is actively monitoring protected roles by analyzing

SQL statements issued by those roles and reacting according to the setting of
edb_sql_protect.level. When you are ready to begin monitoring with SQL/Protect
set this parameter to on.

3.1.3.12.6 edb_sql_protect.level

Parameter Type: Enum

Default Value: passive

Range: {learn | passive | active}

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Reload

Required Authorization to Activate: EPAS service account

Sets the action taken by SQL/Protect when a SQL statement is issued by a protected role.

The edb_sql_protect.level configuration parameter can be set to one of the

following values to use either learn mode, passive mode, or active mode:

learn. Tracks the activities of protected roles and records the relations used by the
roles. This is used when initially configuring SQL/Protect so the expected
behaviors of the protected applications are learned.
passive. Issues warnings if protected roles are breaking the defined rules, but does
not stop any SQL statements from executing. This is the next step after
SQL/Protect has learned the expected behavior of the protected roles. This
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 96
 EDB Postgres Advanced Server Guide

essentially behaves in intrusion detection mode and can be run in production

when properly monitored.
active. Stops all invalid statements for a protected role. This behaves as a SQL
firewall preventing dangerous queries from running. This is particularly effective
against early penetration testing when the attacker is trying to determine the
vulnerability point and the type of database behind the application. Not only does
SQL/Protect close those vulnerability points, but it tracks the blocked queries
allowing administrators to be alerted before the attacker finds an alternate method
of penetrating the system.

If you are using SQL/Protect for the first time, set edb_sql_protect.level to
learn.

3.1.3.12.7 edb_sql_protect.max_protected_relations

Parameter Type: Integer

Default Value: 1024

Range: 1 to 2147483647

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Restart

Required Authorization to Activate: EPAS service account

Sets the maximum number of relations that can be protected per role. Please note the
total number of protected relations for the server will be the number of protected relations
times the number of protected roles. Every protected relation consumes space in shared
memory. The space for the maximum possible protected relations is reserved during
database server startup.

If the server is started when edb_sql_protect.max_protected_relations is set

to a value outside of the valid range (for example, a value of 2,147,483,648), then a
warning message is logged in the database server log file:

2014-07-18 16:04:12 EDT WARNING: invalid value for parameter

"edb_sql_protect.max_protected_relations": "2147483648"
2014-07-18 16:04:12 EDT HINT: Value exceeds integer range.

The database server starts successfully, but with

edb_sql_protect.max_protected_relations set to the default value of 1024.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 97

 EDB Postgres Advanced Server Guide

Though the upper range for the parameter is listed as the maximum value for an integer
data type, the practical setting depends on how much shared memory is available and the
parameter value used during database server startup.

As long as the space required can be reserved in shared memory, the value will be
acceptable. If the value is such that the space in shared memory cannot be reserved, the
database server startup fails with an error message such as the following:

2014-07-18 15:22:17 EDT FATAL: could not map anonymous shared memory: Cannot
allocate memory
2014-07-18 15:22:17 EDT HINT: This error usually means that PostgreSQL's
request for a shared memory segment exceeded available memory, swap space or
huge pages. To reduce the request size (currently 2070118400 bytes), reduce
PostgreSQL's shared memory usage, perhaps by reducing shared_buffers or
max_connections.

In such cases, reduce the parameter value until the database server can be started
successfully.

3.1.3.12.8 edb_sql_protect.max_protected_roles

Parameter Type: Integer

Default Value: 64

Range: 1 to 2147483647

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Restart

Required Authorization to Activate: EPAS service account

Sets the maximum number of roles that can be protected.

Every protected role consumes space in shared memory. Please note that the server will
reserve space for the number of protected roles times the number of protected relations
(edb_sql_protect.max_protected_relations). The space for the maximum
possible protected roles is reserved during database server startup.

If the database server is started when edb_sql_protect.max_protected_roles is

set to a value outside of the valid range (for example, a value of 2,147,483,648), then a
warning message is logged in the database server log file:

2014-07-18 16:04:12 EDT WARNING: invalid value for parameter

"edb_sql_protect.max_protected_roles": "2147483648"
2014-07-18 16:04:12 EDT HINT: Value exceeds integer range.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 98

 EDB Postgres Advanced Server Guide

The database server starts successfully, but with

edb_sql_protect.max_protected_roles set to the default value of 64.

Though the upper range for the parameter is listed as the maximum value for an integer
data type, the practical setting depends on how much shared memory is available and the
parameter value used during database server startup.

As long as the space required can be reserved in shared memory, the value will be
acceptable. If the value is such that the space in shared memory cannot be reserved, the
database server startup fails with an error message such as the following:

2014-07-18 15:22:17 EDT FATAL: could not map anonymous shared memory: Cannot
allocate memory
2014-07-18 15:22:17 EDT HINT: This error usually means that PostgreSQL's
request for a shared memory segment exceeded available memory, swap space or
huge pages. To reduce the request size (currently 2070118400 bytes), reduce
PostgreSQL's shared memory usage, perhaps by reducing shared_buffers or
max_connections.

In such cases, reduce the parameter value until the database server can be started
successfully.

3.1.3.12.9 edb_sql_protect.max_queries_to_save

Parameter Type: Integer

Default Value: 5000

Range: 100 to 2147483647

Minimum Scope of Effect: Cluster

When Value Changes Take Effect: Restart

Required Authorization to Activate: EPAS service account

Sets the maximum number of offending queries to save in view

edb_sql_protect_queries.

Every query that is saved consumes space in shared memory. The space for the maximum
possible queries that can be saved is reserved during database server startup.

If the database server is started when edb_sql_protect.max_queries_to_save is

set to a value outside of the valid range (for example, a value of 10), then a warning
message is logged in the database server log file:

2014-07-18 13:05:31 EDT WARNING: 10 is outside the valid range for parameter
"edb_sql_protect.max_queries_to_save" (100 .. 2147483647)

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 99

 EDB Postgres Advanced Server Guide

The database server starts successfully, but with

edb_sql_protect.max_queries_to_save set to the default value of 5000.

Though the upper range for the parameter is listed as the maximum value for an integer
data type, the practical setting depends on how much shared memory is available and the
parameter value used during database server startup.

As long as the space required can be reserved in shared memory, the value will be
acceptable. If the value is such that the space in shared memory cannot be reserved, the
database server startup fails with an error message such as the following:

2014-07-18 15:22:17 EDT FATAL: could not map anonymous shared memory: Cannot
allocate memory
2014-07-18 15:22:17 EDT HINT: This error usually means that PostgreSQL's
request for a shared memory segment exceeded available memory, swap space or
huge pages. To reduce the request size (currently 2070118400 bytes), reduce
PostgreSQL's shared memory usage, perhaps by reducing shared_buffers or
max_connections.

In such cases, reduce the parameter value until the database server can be started
successfully.

3.1.3.12.10 edbldr.empty_csv_field

Parameter Type: Enum

Default Value: NULL

Range: {NULL | empty_string | pgsql}

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

Use the edbldr.empty_csv_field parameter to specify how EDB*Loader will treat

an empty string. The valid values for the edbldr.empty_csv_field parameter are:

Parameter Setting EDB*Loader Behavior

NULL An empty field is treated as NULL.
empty_string An empty field is treated as a string of length zero.
pgsql An empty field is treated as a NULL if it does not contain quotes and as an empty
string if it contains quotes.

For more information about the edbldr.empty_csv_field parameter in

EDB*Loader, see the Database Compatibility Guide for Oracle Developers.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 100

 EDB Postgres Advanced Server Guide

3.1.3.12.11 utl_encode.uudecode_redwood

Parameter Type: Boolean

Default Value: false

Range: {true | false}

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

When set to TRUE, Advanced Servers UTL_ENCODE.UUDECODE function can decode

uuencoded data that was created by the Oracle implementation of the
UTL_ENCODE.UUENCODE function.

When set to the default setting of FALSE, Advanced Servers UTL_ENCODE.UUDECODE

function can decode uuencoded data created by Advanced Servers
UTL_ENCODE.UUENCODE function.

3.1.3.12.12 utl_file.umask

Parameter Type: String

Default Value: 0077

Range: Octal digits for umask settings

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

The utl_file.umask parameter sets the file mode creation mask or simply, the mask,
in a manner similar to the Linux umask command. This is for usage only within the
Advanced Server UTL_FILE package.

Note: The utl_file.umask parameter is not supported on Windows systems.

The value specified for utl_file.umask is a 3 or 4-character octal string that would be
valid for the Linux umask command. The setting determines the permissions on files
created by the UTL_FILE functions and procedures. (Refer to any information source
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 101
 EDB Postgres Advanced Server Guide

regarding Linux or Unix systems for information on file permissions and the usage of the
umask command.)

The following shows the results of the default utl_file.umask setting of 0077. Note
that all permissions are denied on users belonging to the enterprisedb group as well
as all other users. Only user enterprisedb has read and write permissions on the file.

-rw------- 1 enterprisedb enterprisedb 21 Jul 24 16:08 utlfile

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 102

 EDB Postgres Advanced Server Guide

3.1.3.13 Ungrouped
Configuration parameters in this section apply to Advanced Server only and are for a
specific, limited purpose.

3.1.3.13.1 nls_length_semantics

Parameter Type: Enum

Default Value: byte

Range: {byte | char}

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Superuser

This parameter has no effect in Advanced Server.

For example, the form of the ALTER SESSION command is accepted in Advanced Server
without throwing a syntax error, but does not alter the session environment:

ALTER SESSION SET nls_length_semantics = char;

Note: Since the setting of this parameter has no effect on the server environment, it does
not appear in the system view pg_settings.

3.1.3.13.2 query_rewrite_enabled

Parameter Type: Enum

Default Value: false

Range: {true | false | force}

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 103

 EDB Postgres Advanced Server Guide

This parameter has no effect in Advanced Server.

For example, the following form of the ALTER SESSION command is accepted in
Advanced Server without throwing a syntax error, but does not alter the session
environment:

ALTER SESSION SET query_rewrite_enabled = force;

Note: Since the setting of this parameter has no effect on the server environment, it does
not appear in the system view pg_settings.

3.1.3.13.3 query_rewrite_integrity

Parameter Type: Enum

Default Value: enforced

Range: {enforced | trusted | stale_tolerated}

Minimum Scope of Effect: Per session

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Superuser

This parameter has no effect in Advanced Server.

For example, the following form of the ALTER SESSION command is accepted in
Advanced Server without throwing a syntax error, but does not alter the session
environment:

ALTER SESSION SET query_rewrite_integrity = stale_tolerated;

Note: Since the setting of this parameter has no effect on the server environment, it does
not appear in the system view pg_settings.

3.1.3.13.4 timed_statistics

Parameter Type: Boolean

Default Value: true

Range: {true | false}

Minimum Scope of Effect: Per session

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 104

 EDB Postgres Advanced Server Guide

When Value Changes Take Effect: Immediate

Required Authorization to Activate: Session user

Controls the collection of timing data for the Dynamic Runtime Instrumentation Tools
Architecture (DRITA) feature. When set to on, timing data is collected.

Note: When Advanced Server is installed, the postgresql.conf file contains an

explicit entry setting timed_statistics to off. If this entry is commented out letting
timed_statistics to default, and the configuration file is reloaded, timed statistics
collection would be turned on.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 105

 EDB Postgres Advanced Server Guide

3.2 Index Advisor

The Index Advisor utility helps determine which columns you should index to improve
performance in a given workload. Index Advisor considers B-tree (single-column or
composite) index types, and does not identify other index types (GIN, GiST, Hash) that
may improve performance. Index Advisor is installed with EDB Postgres Advanced
Server.

Index Advisor works with Advanced Server's query planner by creating hypothetical
indexes that the query planner uses to calculate execution costs as if such indexes were
available. Index Advisor identifies the indexes by analyzing SQL queries supplied in the
workload.

There are three ways to use Index Advisor to analyze SQL queries:

Invoke the Index Advisor utility program, supplying a text file containing the
SQL queries that you wish to analyze; Index Advisor will generate a text file with
CREATE INDEX statements for the recommended indexes.
Provide queries at the EDB-PSQL command line that you want Index Advisor to
analyze.
Access Index Advisor through the Postgres Enterprise Manager client. When
accessed via the PEM client, Index Advisor works with SQL Profiler, providing
indexing recommendations on code captured in SQL traces. For more
information about using SQL Profiler and Index Advisor with PEM, please see
the PEM Getting Started Guide available from the EnterpriseDB website at:

http://www.enterprisedb.com/products-services-training/products/postgres-enterprise-
manager

Index Advisor will attempt to make indexing recommendations on INSERT, UPDATE,

DELETE and SELECT statements. When invoking Index Advisor, you supply the
workload in the form of a set of queries (if you are providing the command in an SQL
file) or an EXPLAIN statement (if you are specifying the SQL statement at the psql
command line). Index Advisor displays the query plan and estimated execution cost for
the supplied query, but does not actually execute the query.

During the analysis, Index Advisor compares the query execution costs with and without
hypothetical indexes. If the execution cost using a hypothetical index is less than the
execution cost without it, both plans are reported in the EXPLAIN statement output,
metrics that quantify the improvement are calculated, and Index Advisor generates the
CREATE INDEX statement needed to create the index.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 106

 EDB Postgres Advanced Server Guide

If no hypothetical index can be found that reduces the execution cost, Index Advisor
displays only the original query plan output of the EXPLAIN statement.

Index Advisor does not actually create indexes on the tables. Use the CREATE INDEX
statements supplied by Index Advisor to add any recommended indexes to your tables.

A script supplied with Advanced Server creates the table in which Index Advisor stores
the indexing recommendations generated by the analysis; the script also creates a
function and a view of the table to simplify the retrieval and interpretation of the results.

If you choose to forego running the script, Index Advisor will log recommendations in a
temporary table that is available only for the duration of the Index Advisor session.

3.2.1 Index Advisor Components

The Index Advisor shared library interacts with the query planner to make indexing
recommendations. The Advanced Server installer creates the following shared library in
the libdir subdirectory of your Advanced Server home directory:

On Linux:

index_advisor.so

On Windows:

index_advisor.dll

Please note that libraries in the libdir directory can only be loaded by a superuser. A
database administrator can allow a non-superuser to use Index Advisor by manually
copying the Index Advisor file from the libdir directory into the libdir/plugins
directory (under your Advanced Server home directory). Only a trusted non-superuser
should be allowed access to the plugin; this is an unsafe practice in a production
environment.

The installer also creates the Index Advisor utility program and setup script:

pg_advise_index

pg_advise_index is a utility program that reads a user-supplied input file

containing SQL queries and produces a text file containing CREATE INDEX
statements that can be used to create the indexes recommended by the Index
Advisor. The pg_advise_index program is located in the bin subdirectory of
the Advanced Server home directory.

index_advisor.sql

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 107

 EDB Postgres Advanced Server Guide

index_advisor.sql is a script that creates a permanent Index Advisor log

table along with a function and view to facilitate reporting of recommendations
from the log table. The script is located in the share/contrib subdirectory of
the Advanced Server directory.

The index_advisor.sql script creates the index_advisor_log table, the

show_index_recommendations() function and the index_recommendations
view. These database objects must be created in a schema that is accessible by, and
included in the search path of the role that will invoke Index Advisor.

index_advisor_log

Index Advisor logs indexing recommendations in the index_advisor_log

table. If Index Advisor does not find the index_advisor_log table in the
user's search path, Index Advisor will store any indexing recommendations in a
temporary table of the same name. The temporary table exists only for the
duration of the current session.

show_index_recommendations()

show_index_recommendations() is a PL/pgSQL function that interprets and

displays the recommendations made during a specific Index Advisor session (as
identified by its backend process ID).

index_recommendations

Index Advisor creates the index_recommendations view based on information

stored in the index_advisor_log table during a query analysis. The view
produces output in the same format as the show_index_recommendations()
function, but contains Index Advisor recommendations for all stored sessions,
while the result set returned by the show_index_recommendations()
function are limited to a specified session.

3.2.2 Index Advisor Configuration

Index Advisor does not require any configuration to generate recommendations that are
available only for the duration of the current session; to store the results of multiple
sessions, you must create the index_advisor_log table (where Advanced Server will
store Index Advisor recommendations). To create the index_advisor_log table , you
must run the index_advisor.sql script.

When selecting a storage schema for the Index Advisor table, function and view, keep in
mind that all users that invoke Index Advisor (and query the result set) must have USAGE

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 108

 EDB Postgres Advanced Server Guide

privileges on the schema. The schema must be in the search path of all users that are
interacting with the Index Advisor.

1. Place the selected schema at the start of your search_path parameter. For
example, if your search path is currently:

search_path=public, accounting
and you want the Index Advisor objects to be created in a schema named
advisor, use the command:
SET search_path = advisor, public, accounting;

2. Run the index_advisor.sql script to create the database objects. If you

are running the psql client, you can use the command:

\i full_pathname/index_advisor.sql
Specify the pathname to the index_advisor.sql script in place of
full_pathname.

3. Grant privileges on the index_advisor_log table to all Index Advisor

users; this step is not necessary if the Index Advisor user is a superuser, or the
owner of these database objects.

Grant SELECT and INSERT privileges on the index_advisor_log table

to allow a user to invoke Index Advisor.

Grant DELETE privileges on the index_advisor_log table to allow the

specified user to delete the table contents.

Grant SELECT privilege on the index_recommendations view.

The following example demonstrates the creation of the Index Advisor database objects
in a schema named ia, which will then be accessible to an Index Advisor user with user
name ia_user:

$ edb-psql -d edb -U enterprisedb

edb-psql (10.0.1)
Type "help" for help.

edb=# CREATE SCHEMA ia;

CREATE SCHEMA
edb=# SET search_path TO ia;
SET
edb=# \i /opt/edb/as10/share/contrib/index_advisor.sql
CREATE TABLE
CREATE INDEX
CREATE INDEX
CREATE FUNCTION
CREATE FUNCTION
CREATE VIEW
edb=# GRANT USAGE ON SCHEMA ia TO ia_user;
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 109
 EDB Postgres Advanced Server Guide

GRANT
edb=# GRANT SELECT, INSERT, DELETE ON index_advisor_log TO ia_user;
GRANT
edb=# GRANT SELECT ON index_recommendations TO ia_user;
GRANT

While using Index Advisor, the specified schema (ia) must be included in ia_user's
search_path parameter.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 110

 EDB Postgres Advanced Server Guide

3.2.3 Using Index Advisor

When you invoke Index Advisor, you must supply a workload; the workload is either a
query (specified at the command line), or a file that contains a set of queries (executed by
the pg_advise_index() function). After analyzing the workload, Index Advisor will
either store the result set in a temporary table, or in a permanent table. You can review
the indexing recommendations generated by Index Advisor and use the CREATE INDEX
statements generated by Index Advisor to create the recommended indexes.

Note: You should not run Index Advisor in read-only transactions.

The following examples assume that superuser enterprisedb is the Index Advisor
user, and the Index Advisor database objects have been created in a schema in the
search_path of superuser enterprisedb.

The examples in the following sections use the table created with the statement shown
below:

CREATE TABLE t( a INT, b INT );

INSERT INTO t SELECT s, 99999 - s FROM generate_series(0,99999) AS s;
ANALYZE t;

The resulting table contains the following rows:

a | b
-------+-------
0 | 99999
1 | 99998
2 | 99997
3 | 99996
.
.
.
99997 | 2
99998 | 1
99999 | 0

3.2.3.1 Using the pg_advise_index Utility

When invoking the pg_advise_index utility, you must include the name of a file that
contains the queries that will be executed by pg_advise_index; the queries may be on
the same line, or on separate lines, but each query must be terminated by a semicolon.
Queries within the file should not begin with the EXPLAIN keyword.

The following example shows the contents of a sample workload.sql file:

SELECT * FROM t WHERE a = 500;

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 111
 EDB Postgres Advanced Server Guide

SELECT * FROM t WHERE b < 1000;

Run the pg_advise_index program as shown in the code sample below:

$ pg_advise_index -d edb -h localhost -U enterprisedb -s 100M -o advisory.sql

workload.sql
poolsize = 102400 KB
load workload from file 'workload.sql'
Analyzing queries .. done.
size = 2184 KB, benefit = 1684.720000
size = 2184 KB, benefit = 1655.520000
/* 1. t(a): size=2184 KB, benefit=1684.72 */
/* 2. t(b): size=2184 KB, benefit=1655.52 */
/* Total size = 4368KB */

In the code sample, the -d, -h, and -U options are psql connection options.

-s

-s is an optional parameter that limits the maximum size of the indexes

recommended by Index Advisor. If Index Advisor does not return a result set, -s
may be set too low.

-o

The recommended indexes are written to the file specified after the -o option.

The information displayed by the pg_advise_index program is logged in the

index_advisor_log table. In response to the command shown in the example, Index
Advisor writes the following CREATE INDEX statements to the advisory.sql output
file

create index idx_t_1 on t (a);

create index idx_t_2 on t (b);

You can create the recommended indexes at the psql command line with the CREATE
INDEX statements in the file, or create the indexes by executing the advisory.sql
script.

$ edb-psql -d edb -h localhost -U enterprisedb -e -f advisory.sql

create index idx_t_1 on t (a);
CREATE INDEX
create index idx_t_2 on t (b);
CREATE INDEX

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 112

 EDB Postgres Advanced Server Guide

3.2.3.2 Using Index Advisor at the psql Command Line

You can use Index Advisor to analyze SQL statements entered at the edb-psql (or psql)
command line; the following steps detail loading the Index Advisor plugin and using
Index Advisor:

1. Connect to the server with the edb-psql command line utility, and load the Index
Advisor plugin:

$ edb-psql -d edb -U enterprisedb

edb=# LOAD 'index_advisor';
LOAD

2. Use the edb-psql command line to invoke each SQL command that you would like
Index Advisor to analyze. Index Advisor stores any recommendations for the queries
in the index_advisor_log table. If the index_advisor_log table does not exist
in the user's search_path, a temporary table is created with the same name. This
temporary table exists only for the duration of the user's session.

After loading the Index Advisor plugin, Index Advisor will analyze all SQL statements
and log any indexing recommendations for the duration of the session.

If you would like Index Advisor to analyze a query (and make indexing
recommendations) without actually executing the query, preface the SQL
statement with the EXPLAIN keyword.

If you do not preface the statement with the EXPLAIN keyword, Index Advisor
will analyze the statement while the statement executes, writing the indexing
recommendations to the index_advisor_log table for later review.

In the example that follows, the EXPLAIN statement displays the normal query plan,
followed by the query plan of the same query, if the query were using the recommended
hypothetical index:

edb=# EXPLAIN SELECT * FROM t WHERE a < 10000;

QUERY PLAN
-----------------------------------------------------------------------------
Seq Scan on t (cost=0.00..1693.00 rows=10105 width=8)
Filter: (a < 10000)
Result (cost=0.00..337.10 rows=10105 width=8)
One-Time Filter: '===[ HYPOTHETICAL PLAN ]==='::text
-> Index Scan using "<hypothetical-index>:1" on t
(cost=0.00..337.10 rows=10105 width=8)
Index Cond: (a < 10000)
(6 rows)

edb=# EXPLAIN SELECT * FROM t WHERE a = 100;

QUERY PLAN
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 113
 EDB Postgres Advanced Server Guide

-----------------------------------------------------------------------------
Seq Scan on t (cost=0.00..1693.00 rows=1 width=8)
Filter: (a = 100)
Result (cost=0.00..8.28 rows=1 width=8)
One-Time Filter: '===[ HYPOTHETICAL PLAN ]==='::text
-> Index Scan using "<hypothetical-index>:3" on t
(cost=0.00..8.28 rows=1 width=8)
Index Cond: (a = 100)
(6 rows)

After loading the Index Advisor plugin, the default value of index_advisor.enabled
is on. The Index Advisor plugin must be loaded to use a SET or SHOW command to
display the current value of index_advisor.enabled.

You can use the index_advisor.enabled parameter to temporarily disable Index

Advisor without interrupting the psql session:

edb=# SET index_advisor.enabled TO off;

SET

To enable Index Advisor, set the parameter to on:

edb=# SET index_advisor.enabled TO on;

SET

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 114

 EDB Postgres Advanced Server Guide

3.2.4 Reviewing the Index Advisor Recommendations

There are several ways to review the index recommendations generated by Index
Advisor. You can:

Query the index_advisor_log table.

Run the show_index_recommendations function.
Query the index_recommendations view.

3.2.4.1 Using the show_index_recommendations() Function

To review the recommendations of the Index Advisor utility using the
show_index_recommendations() function, call the function, specifying the process
ID of the session:

SELECT show_index_recommendations( pid );

Where pid is the process ID of the current session. If you do not know the process ID of
your current session, passing a value of NULL will also return a result set for the current
session.

The following code fragment shows an example of a row in a result set:

edb=# SELECT show_index_recommendations(null);

show_index_recommendations
---------------------------------------------------------------------
create index idx_t_a on t(a);/* size: 2184 KB, benefit: 3040.62,
gain: 1.39222666981456 */
(1 row)

In the example, create index idx_t_a on t(a) is the SQL statement needed to create
the index suggested by Index Advisor. Each row in the result set shows:

The command required to create the recommended index.

The maximum estimated size of the index.
The calculated benefit of using the index.
The estimated gain that will result from implementing the index.

You can display the results of all Index Advisor sessions from the following view:

SELECT * FROM index_recommendations;

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 115

 EDB Postgres Advanced Server Guide

3.2.4.2 Querying the index_advisor_log Table

Index Advisor stores indexing recommendations in a table named
index_advisor_log. Each row in the index_advisor_log table contains the result
of a query where Index Advisor determines it can recommend a hypothetical index to
reduce the execution cost of that query.

Column Type Description

reloid oid OID of the base table for the index
relname name Name of the base table for the index
attrs integer[] Recommended index columns (identified by column number)
benefit real Calculated benefit of the index for this query
index_size integer Estimated index size in disk-pages
backend_pid integer Process ID of the process generating this recommendation
timestamp timestamp Date/Time when the recommendation was generated

You can query the index_advisor_log table at the psql command line. The following
example shows the index_advisor_log table entries resulting from two Index
Advisor sessions. Each session contains two queries, and can be identified (in the table
below) by a different backend_pid value. For each session, Index Advisor generated
two index recommendations.

edb=# SELECT * FROM index_advisor_log;

reloid | relname | attrs | benefit | index_size | backend_pid |
timestamp
--------+---------+-------+---------+------------+-------------+-----------
-----------------------
16651 | t | {1} | 1684.72 | 2184 | 3442 | 22-MAR-11
16:44:32.712638 -04:00
16651 | t | {2} | 1655.52 | 2184 | 3442 | 22-MAR-11
16:44:32.759436 -04:00
16651 | t | {1} | 1355.9 | 2184 | 3506 | 22-MAR-11
16:48:28.317016 -04:00
16651 | t | {1} | 1684.72 | 2184 | 3506 | 22-MAR-11
16:51:45.927906 -04:00
(4 rows)

Index Advisor added the first two rows to the table after analyzing the following two
queries executed by the pg_advise_index utility:

SELECT * FROM t WHERE a = 500;

SELECT * FROM t WHERE b < 1000;

The value of 3442 in column backend_pid identifies these results as coming from the
session with process ID 3442.

The value of 1 in column attrs in the first row indicates that the hypothetical index is
on the first column of the table (column a of table t).

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 116

 EDB Postgres Advanced Server Guide

The value of 2 in column attrs in the second row indicates that the hypothetical index
is on the second column of the table (column b of table t).

Index Advisor added the last two rows to the table after analyzing the following two
queries (executed at the psql command line):

edb=# EXPLAIN SELECT * FROM t WHERE a < 10000;

QUERY PLAN
--------------------------------------------------------------------------
--------------------
Seq Scan on t (cost=0.00..1693.00 rows=10105 width=8)
Filter: (a < 10000)
Result (cost=0.00..337.10 rows=10105 width=8)
One-Time Filter: '===[ HYPOTHETICAL PLAN ]==='::text
-> Index Scan using "<hypothetical-index>:1" on t (cost=0.00..337.10
rows=10105 width=8)
Index Cond: (a < 10000)
(6 rows)

edb=# EXPLAIN SELECT * FROM t WHERE a = 100;

QUERY PLAN
--------------------------------------------------------------------------
--------------
Seq Scan on t (cost=0.00..1693.00 rows=1 width=8)
Filter: (a = 100)
Result (cost=0.00..8.28 rows=1 width=8)
One-Time Filter: '===[ HYPOTHETICAL PLAN ]==='::text
-> Index Scan using "<hypothetical-index>:3" on t (cost=0.00..8.28
rows=1 width=8)
Index Cond: (a = 100)
(6 rows)

The values in the benefit column of the index_advisor_log table are calculated using
the following formula:

benefit = (normal execution cost) - (execution cost with hypothetical

index)

The value of the benefit column for the last row of the index_advisor_log table
(shown in the example) is calculated using the query plan for the following SQL
statement:

EXPLAIN SELECT * FROM t WHERE a = 100;

The execution costs of the different execution plans are evaluated and compared:

benefit = (Seq Scan on t cost) - (Index Scan using

<hypothetical-index>)

and the benefit is added to the table:

benefit = 1693.00 - 8.28

benefit = 1684.72
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 117
 EDB Postgres Advanced Server Guide

You can delete rows from the index_advisor_log table when you no longer have the
need to review the results of the queries stored in the row.

3.2.4.3 Querying the index_recommendations View

The index_recommendations view contains the calculated metrics and the CREATE
INDEX statements to create the recommended indexes for all sessions whose results are
currently in the index_advisor_log table. You can display the results of all stored
Index Advisor sessions by querying the index_recommendations view as shown
below:

SELECT * FROM index_recommendations;

Using the example shown in the previous section (Querying the index_advisor_log
Table), the index_recommendations view displays the following:

edb=# SELECT * FROM index_recommendations;

backend_pid | show_index_recommendations
-------------+-------------------------------------------------------------
--------------------------------
3442 | create index idx_t_a on t(a);/* size: 2184 KB, benefit:
1684.72, gain: 0.771392654586624 */
3442 | create index idx_t_b on t(b);/* size: 2184 KB, benefit:
1655.52, gain: 0.758021539820856 */
3506 | create index idx_t_a on t(a);/* size: 2184 KB, benefit:
3040.62, gain: 1.39222666981456 */
(3 rows)

Within each session, the results of all queries that benefit from the same recommended
index are combined to produce one set of metrics per recommended index, reflected in
the fields named benefit and gain.

The formulas for the fields are as follows:

size = MAX(index size of all queries)

benefit = SUM(benefit of each query)
gain = SUM(benefit of each query) / MAX(index size of all
queries)

So for example, using the following query results from the process with a backend_pid
of 3506:

reloid | relname | attrs | benefit | index_size | backend_pid |

timestamp
--------+---------+-------+---------+------------+-------------+-----------
-----------------------
16651 | t | {1} | 1355.9 | 2184 | 3506 | 22-MAR-11
16:48:28.317016 -04:00
16651 | t | {1} | 1684.72 | 2184 | 3506 | 22-MAR-11
16:51:45.927906 -04:00

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 118

 EDB Postgres Advanced Server Guide

The metrics displayed from the index_recommendations view for backend_pid

3506 are:

backend_pid | show_index_recommendations
-------------+-------------------------------------------------------------
--------------------------------
3506 | create index idx_t_a on t(a);/* size: 2184 KB, benefit:
3040.62, gain: 1.39222666981456 */

The metrics from the view are calculated as follows:

benefit = (benefit from 1st query) + (benefit from 2nd query)

benefit = 1355.9 + 1684.72
benefit = 3040.62

and

gain = ((benefit from 1st query) + (benefit from 2nd query))

/ MAX(index size of all queries)
gain = (1355.9 + 1684.72) / MAX(2184, 2184)
gain = 3040.62 / 2184
gain = 1.39223

The gain metric is useful when comparing the relative advantage of the different
recommended indexes derived during a given session. The larger the gain value, the
better the cost effectiveness derived from the index weighed against the possible disk
space consumption of the index.

3.2.5 Limitations

Index Advisor does not consider Index Only scans; it does consider Index scans when
making recommendations.

Index Advisor ignores any computations found in the WHERE clause. Effectively, the
index field in the recommendations will not be any kind of expression; the field will be a
simple column name.

Index Advisor does not consider inheritance when recommending hypothetical indexes.
If a query references a parent table, Index Advisor does not make any index
recommendations on child tables.

Restoration of a pg_dump backup file that includes the index_advisor_log table or

any tables for which indexing recommendations were made and stored in the
index_advisor_log table, may result in "broken links" between the
index_advisor_log table and the restored tables referenced by rows in the
index_advisor_log table because of changes in object identifiers (OIDs).

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 119

 EDB Postgres Advanced Server Guide

If it is necessary to display the recommendations made prior to the backup, you can
replace the old OIDs in the reloid column of the index_advisor_log table with the
new OIDs of the referenced tables using the SQL UPDATE statement:

UPDATE index_advisor_log SET reloid = new_oid WHERE reloid =

old_oid;

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 120

 EDB Postgres Advanced Server Guide

3.3 SQL Profiler

Inefficient SQL code is one of, if not the leading cause of database performance
problems. The challenge for database administrators and developers is locating and then
optimizing this code in large, complex systems.

SQL Profiler helps you locate and optimize poorly running SQL code.

Specific features and benefits of SQL Profiler include the following:

On-Demand Traces. You can capture SQL traces at any time by manually
setting up your parameters and starting the trace.
Scheduled Traces. For inconvenient times, you can also specify your trace
parameters and schedule them to run at some later time.
Save Traces. Execute your traces and save them for later review.
Trace Filters. Selectively filter SQL captures by database and by user, or capture
every SQL statement sent by all users against all databases.
Trace Output Analyzer. A graphical table lets you quickly sort and filter queries
by duration or statement, and a graphical or text based EXPLAIN plan lays out
your query paths and joins.
Index Advisor Integration. Once you have found your slow queries and
optimized them, you can also let the Index Advisor recommend the creation of
underlying table indices to further improve performance.

The following describes the installation process.

Step 1: Install SQL Profiler

SQL Profiler is installed by the Advanced Server Installer, or you can download and
install SQL Profiler into a managed database instance.

Step 2: Add the SQL Profiler library

Modify the postgresql.conf parameter file for the instance to include the SQL
Profiler library in the shared_preload_libraries configuration parameter.

For Linux installations, the parameter value should include:

$libdir/sql-profiler

On Windows, the parameter value should include:

$libdir\sql-profiler.dll

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 121

 EDB Postgres Advanced Server Guide

Step 3: Create the functions used by SQL Profiler

The SQL Profiler installation program places a SQL script (named sql-
profiler.sql) in:

On Linux:

/opt/edb/as10/share/contrib/

On Windows:

C:\Program Files\edb\as10\share\contrib\

Use the psql command line interface to run the sql-profiler.sql script in the
database specified as the Maintenance Database on the server you wish to profile. If you
are using Advanced Server, the default maintenance database is named edb. If you are
using a PostgreSQL instance, the default maintenance database is named postgres.

The following command uses the psql command line to invoke the sql-
profiler.sql script on a Linux system:

$ /opt/edb/as10/bin/psql -U user_name database_name <

/opt/edb/as10/share/contrib/sql-profiler.sql

Step 4: Stop and restart the server for the changes to take effect.

After configuring SQL Profiler, it is ready to use with all databases that reside on the
server. You can take advantage of SQL Profiler functionality with EDB Postgres
Enterprise Manager; for more information about Postgres Enterprise Manager, visit the
EnterpriseDB website at:

http://www.enterprisedb.com/products/postgres-enterprise-manager

Troubleshooting

If (after performing an upgrade to a newer version of SQL Profiler) you encounter an

error that contains the following text:

An error has occurred:

ERROR: function return row and query-specified return row do not match.
DETAIL: Returned row contains 11 attributes, but the query expects 10.

To correct this error, you must replace the existing query set with a new query set. First,
uninstall SQL Profiler by invoking the uninstall-sql-profiler.sql script, and
then reinstall SQL Profiler by invoking the sql-profiler.sql script.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 122

 EDB Postgres Advanced Server Guide

3.4 pgsnmpd

pgsnmpd is an SNMP agent that can return hierarchical information about the current
state of Advanced Server on a Linux system. pgsnmpd is distributed with and installed
by the Advanced Server installer as part of the database server component. The pgsnmpd
agent can operate as a stand-alone SNMP agent, as a pass-through sub-agent, or as an
AgentX sub-agent.

After installing Advanced Server, you will need to update the LD_LIBRARY_PATH
variable. Use the command:

$ export LD_LIBRARY_PATH=/opt/edb/as10/lib:$LD_LIBRARY_PATH

This command does not persistently alter the value of LD_LIBRARY_PATH. Consult the
documentation for your distribution of Linux for information about persistently setting
the value of LD_LIBRARY_PATH.

The examples that follow demonstrate the simplest usage of pgsnmpd, implementing
read only access. pgsnmpd is based on the net-snmp library; for more information about
net-snmp, visit:

http://net-snmp.sourceforge.net/

3.4.1 Configuring pgsnmpd

The pgsnmpd configuration file is named snmpd.conf. For information about the
directives that you can specify in the configuration file, please review the snmpd.conf
man page (man snmpd.conf).

You can create the configuration file by hand, or you can use the snmpconf perl script to
create the configuration file. The perl script is distributed with net-snmp package.

net-snmp is an open-source package available from:

http://www.net-snmp.org/

To use the snmpconf configuration file wizard, download and install net-snmp. When the
installation completes, open a command line and enter:

snmpconf

When the configuration file wizard opens, it may prompt you to read in an existing
configuration file. Enter none to generate a new configuration file (not based on a
previously existing configuration file).

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 123

 EDB Postgres Advanced Server Guide

snmpconf is a menu-driven wizard. Select menu item 1: snmpd.conf to start the

configuration wizard. As you select each top-level menu option displayed by snmpconf,
the wizard walks through a series of questions, prompting you for information required to
build the configuration file. When you have provided information in each of the category
relevant to your system, enter Finished to generate a configuration file named
snmpd.conf. Copy the file to:

/opt/edb/as10/share/

3.4.2 Setting the Listener Address

By default, pgsnmpd listens on port 161. If the listener port is already being used by
another service, you may receive the following error:

Error opening specified endpoint "udp:161".

You can specify an alternate listener port by adding the following line to your
snmpd.conf file:

agentaddress $host_address:2000

The example instructs pgsnmpd to listen on UDP port 2000, where $host_address is
the IP address of the server (e.g., 127.0.0.1).

3.4.3 Invoking pgsnmpd

Ensure that an instance of Advanced Server is up and running (pgsnmpd will connect to
this server). Open a command line and assume superuser privileges, before invoking
pgsnmpd with a command that takes the following form:

POSTGRES_INSTALL_HOME/bin/pgsnmpd -b
-c POSTGRES_INSTALL_HOME/share/snmpd.conf
-C "user=enterprisedb dbname=edb password=safe_password
port=5444"

Where POSTGRES_INSTALL_HOME specifies the Advanced Server installation directory.

Include the -b option to specify that pgsnmpd should run in the background.

Include the -c option, specifying the path and name of the pgsnmpd configuration file.

Include connection information for your installation of Advanced Server (in the form of a
libpq connection string) after the -C option.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 124

 EDB Postgres Advanced Server Guide

3.4.4 Viewing pgsnmpd Help

Include the --help option when invoking the pgsnmpd utility to view other pgsnmpd
command line options:
pgsnmpd --help
Version PGSQL-SNMP-Ver1.0
usage: pgsnmpd [-s] [-b] [-c FILE ] [-x address ] [-g] [-C "Connect String"]
-s : run as AgentX sub-agent of an existing snmpd process
-b : run in the background
-c : configuration file name
-g : use syslog
-C : libpq connection string
-x : address:port of a network interface
-V : display version strings

3.4.5 Requesting Information from pgsnmpd

You can use net-snmp commands to query the pgsnmpd service. For example:

snmpgetnext -v 2c -c public localhost

.1.3.6.1.4.1.5432.1.4.2.1.1.0

In the above example:

-v 2c option instructs the snmpgetnext client to send the request in SNMP

version 2c format.

-c public specifies the community name.

localhost indicates the host machine running the pgsnmpd server.

.1.3.6.1.4.1.5432.1.4.2.1.1.0 is the identity of the requested object. To

see a list of all databases, increment the last digit by one (e.g. .1.1, .1.2, .1.3 etc.).

The encodings required to query any given object are defined in the MIB (Management
Information Base). An SNMP client can monitor a variety of servers; the server type
determines the information exposed by a given server. Each SNMP server describes the
exposed data in the form of a MIB (Management information base). By default, pgsnmpd
searches for MIB's in the following locations:

/usr/share/snmp/mibs

$HOME/.snmp/mibs

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 125

 EDB Postgres Advanced Server Guide

3.5 EDB Audit Logging

Advanced Server allows database and security administrators, auditors, and operators to
track and analyze database activities using the EDB Audit Logging functionality. EDB
Audit Logging generates audit log files, which contains all of the relevant information.
The audit logs can be configured to record information such as:

When a role establishes a connection to an Advanced Server database

What database objects a role creates, modifies, or deletes when connected to
Advanced Server
When any failed authentication attempts occur

The parameters specified in the configuration files, postgresql.conf or

postgresql.auto.conf, control the information included in the audit logs.

3.5.1 Audit Logging Configuration Parameters

Use the following configuration parameters to control database auditing. See Section
3.1.2 to determine if a change to the configuration parameter takes effect immediately, or
if the configuration needs to be reloaded, or if the database server needs to be restarted.

edb_audit

Enables or disables database auditing. The values xml or csv will enable
database auditing. These values represent the file format in which auditing
information will be captured. none will disable database auditing and is also the
default.

edb_audit_directory

Specifies the directory where the log files will be created. The path of the
directory can be relative or absolute to the data folder. The default is the
PGDATA/edb_audit directory.

edb_audit_filename

Specifies the file name of the audit file where the auditing information will be
stored. The default file name will be audit-%Y%m%d_%H%M%S. The escape
sequences, %Y, %m etc., will be replaced by the appropriate current values
according to the system date and time.

edb_audit_rotation_day
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 126
 EDB Postgres Advanced Server Guide

Specifies the day of the week on which to rotate the audit files. Valid values are
sun, mon, tue, wed, thu, fri, sat, every, and none. To disable rotation, set
the value to none. To rotate the file every day, set the
edb_audit_rotation_day value to every. To rotate the file on a specific day
of the week, set the value to the desired day of the week. every is the default
value.

edb_audit_rotation_size

Specifies a file size threshold in megabytes when file rotation will be forced to
occur. The default value is 0 MB. If the parameter is commented out or set to 0,
rotation of the file on a size basis will not occur.

edb_audit_rotation_seconds

Specifies the rotation time in seconds when a new log file should be created. To
disable this feature, set this parameter to 0, which is the default.

edb_audit_connect

Enables auditing of database connection attempts by users. To disable auditing of

all connection attempts, set edb_audit_connect to none. To audit all failed
connection attempts, set the value to failed, which is the default. To audit all
connection attempts, set the value to all.

edb_audit_disconnect

Enables auditing of database disconnections by connected users. To enable

auditing of disconnections, set the value to all. To disable, set the value to none,
which is the default.

edb_audit_statement

This configuration parameter is used to specify auditing of different categories of

SQL statements. Various combinations of the following values may be specified:
none, dml, insert, update, delete, truncate, select, error, rollback,
ddl, create, drop, alter, grant, revoke, and all. The default is ddl and
error. See Section 3.5.2 for information regarding the setting of this parameter.

edb_audit_tag

Use this configuration parameter to specify a string value that will be included in
the audit log file for each entry as a tracking tag.

edb_log_every_bulk_value

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 127

 EDB Postgres Advanced Server Guide

Bulk processing logs the resulting statements into both the Advanced Server log
file and the EDB Audit log file. However, logging each and every statement in
bulk processing is costly. This can be controlled by the
edb_log_every_bulk_value configuration parameter. When set to true,
each and every statement in bulk processing is logged. When set to false, a log
message is recorded once per bulk processing. In addition, the duration is emitted
once per bulk processing. Default is false.

edb_audit_destination

Specifies whether the audit log information is to be recorded in the directory as

given by the edb_audit_directory parameter or to the directory and file
managed by the syslog process. Set to file to use the directory specified by
edb_audit_directory, which is the default setting. Set to syslog to use the
syslog process and its location as configured in the /etc/syslog.conf file.
Note: In recent Linux versions, syslog has been replaced by rsyslog and the
configuration file is in /etc/rsyslog.conf.

The following section describes selection of specific SQL statements for auditing using
the edb_audit_statement parameter.

3.5.2 Selecting SQL Statements to Audit

The edb_audit_statement permits inclusion of one or more, comma-separated values

to control which SQL statements are to be audited. The following is the general format:

edb_audit_statement = 'value_1[, value_2]...'

The comma-separated values may include or omit space characters following the comma.
The values can be specified in any combination of lowercase or uppercase letters.

The basic parameter values are the following:

all Results in the auditing and logging of every statement including any error
messages on statements.
none Disables all auditing and logging. A value of none overrides any other
value included in the list.
ddl Results in the auditing of all data definition language (DDL) statements
(CREATE, ALTER, and DROP) as well as GRANT and REVOKE data control language
(DCL) statements.
dml Results in the auditing of all data manipulation language (DML) statements
(INSERT, UPDATE, DELETE, and TRUNCATE).
select Results in the auditing of SELECT statements.
rollback Results in the auditing of ROLLBACK statements.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 128

 EDB Postgres Advanced Server Guide

error Results in the logging of all error messages that occur. Unless the error
value is included, no error messages are logged regarding any errors that occur on
SQL statements related to any of the other preceding parameter values except
when all is used.

Section 3.5.2.1 describes additional parameter values for selecting particular DDL or
DCL statements for auditing.

Section 3.5.2.2 describes additional parameter values for selecting particular DML
statements for auditing.

If an unsupported value is included in the edb_audit_statement parameter, then an

error occurs when applying the setting to the database server. See the database server log
file for the error such as in the following example where create materialized vw
results in the error. (The correct value is create materialized view.)
2017-07-16 11:20:39 EDT LOG: invalid value for parameter "edb_audit_statement": "create
materialized vw, create sequence, grant"
2017-07-16 11:20:39 EDT FATAL: configuration file "/opt/edb/as10/data/postgresql.conf"
contains errors

The following sections describe the values for the SQL language types DDL, DCL, and
DML.

3.5.2.1 Data Definition Language and Data Control Language

Statements

This section describes values that can be included in the edb_audit_statement

parameter to audit DDL and DCL statements.

The following general rules apply:

If the edb_audit_statement parameter includes either ddl or all, then all

DDL statements are audited. In addition, the DCL statements GRANT and REVOKE
are audited as well.
If the edb_audit_statement is set to none, then no DDL nor DCL statements
are audited.
Specific types of DDL and DCL statements can be chosen for auditing by
including a combination of values within the edb_audit_statement
parameter.

Use the following syntax to specify an edb_audit_statement parameter value for

DDL statements:

{ create | alter | drop } [ object_type ]

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 129

 EDB Postgres Advanced Server Guide

object_type is any of the following:

ACCESS METHOD
AGGREGATE
CAST
COLLATION
CONVERSION
DATABASE
EVENT TRIGGER
EXTENSION
FOREIGN TABLE
FUNCTION
INDEX
LANGUAGE
LARGE OBJECT
MATERIALIZED VIEW
OPERATOR
OPERATOR CLASS
OPERATOR FAMILY
POLICY
PUBLICATION
ROLE
RULE
SCHEMA
SEQUENCE
SERVER
SUBSCRIPTION
TABLE
TABLESPACE
TEXT SEARCH CONFIGURATION
TEXT SEARCH DICTIONARY
TEXT SEARCH PARSER
TEXT SEARCH TEMPLATE
TRANSFORM
TRIGGER
TYPE
USER MAPPING
VIEW

Descriptions of object types as used in SQL commands can be found in the PostgreSQL
core documentation available at:

https://www.postgresql.org/docs/10/static/sql-commands.html

If object_type is omitted from the parameter value, then all of the specified command
statements (either create, alter, or drop) are audited.

Use the following syntax to specify an edb_audit_statement parameter value for

DCL statements:
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 130
 EDB Postgres Advanced Server Guide

{ grant | revoke }

The following are some DDL and DCL examples.

Example 1

The following is an example where edb_audit_connect and

edb_audit_statement are set with the following non-default values:

edb_audit_connect = 'all'
edb_audit_statement = 'create, alter, error'

Thus, only SQL statements invoked by the CREATE and ALTER commands are audited.
Error messages are also included in the audit log.

The database session that occurs is the following:

$ psql edb enterprisedb
Password for user enterprisedb:
psql.bin (10.0.1)
Type "help" for help.

edb=# SHOW edb_audit_connect;

edb_audit_connect
-------------------
all
(1 row)

edb=# SHOW edb_audit_statement;

edb_audit_statement
----------------------
create, alter, error
(1 row)

edb=# CREATE ROLE adminuser;

CREATE ROLE
edb=# ALTER ROLE adminuser WITH LOGIN, SUPERUSER, PASSWORD 'password';
ERROR: syntax error at or near ","
LINE 1: ALTER ROLE adminuser WITH LOGIN, SUPERUSER, PASSWORD 'passwo...
^
edb=# ALTER ROLE adminuser WITH LOGIN SUPERUSER PASSWORD 'password';
ALTER ROLE
edb=# CREATE DATABASE auditdb;
CREATE DATABASE
edb=# ALTER DATABASE auditdb OWNER TO adminuser;
ALTER DATABASE
edb=# \c auditdb adminuser
Password for user adminuser:
You are now connected to database "auditdb" as user "adminuser".
auditdb=# CREATE SCHEMA edb;
CREATE SCHEMA
auditdb=# SET search_path TO edb;
SET
auditdb=# CREATE TABLE department (
auditdb(# deptno NUMBER(2) NOT NULL CONSTRAINT dept_pk PRIMARY KEY,
auditdb(# dname VARCHAR2(14) CONSTRAINT dept_dname_uq UNIQUE,
auditdb(# loc VARCHAR2(13)
auditdb(# );
CREATE TABLE
auditdb=# DROP TABLE department;
DROP TABLE
auditdb=# CREATE TABLE dept (

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 131

 EDB Postgres Advanced Server Guide
auditdb(# deptno NUMBER(2) NOT NULL CONSTRAINT dept_pk PRIMARY KEY,
auditdb(# dname VARCHAR2(14) CONSTRAINT dept_dname_uq UNIQUE,
auditdb(# loc VARCHAR2(13)
auditdb(# );
CREATE TABLE

The resulting audit log file contains the following.

Each audit log entry has been split and displayed across multiple lines, and a blank line
has been inserted between the audit log entries for more clarity in the appearance of the
results.
2017-07-16 12:59:42.125 EDT,"enterprisedb","edb",3356,"[local]",
596b9b7e.d1c,1,"authentication",2017-07-16 12:59:42 EDT,6/2,0,AUDIT,00000,
"connection authorized: user=enterprisedb database=edb",,,,,,,,,"","",""

2017-07-16 12:59:42.125 EDT,"enterprisedb","edb",3356,"[local]",

596b9b7e.d1c,2,"idle",2017-07-16 12:59:42 EDT,6/6,0,AUDIT,00000,
"statement: CREATE ROLE adminuser;",,,,,,,,,"psql.bin","CREATE ROLE",""

2017-07-16 13:00:28.469 EDT,"enterprisedb","edb",3356,"[local]",

596b9b7e.d1c,3,"idle",2017-07-16 12:59:42 EDT,6/7,0,ERROR,42601,
"syntax error at or near "",""",,,,,,
"ALTER ROLE adminuser WITH LOGIN, SUPERUSER, PASSWORD 'password';",32,,"psql.bin","",""

2017-07-16 13:00:28.469 EDT,"enterprisedb","edb",3356,"[local]",

596b9b7e.d1c,4,"idle",2017-07-16 12:59:42 EDT,6/8,0,AUDIT,00000,
"statement: ALTER ROLE adminuser WITH LOGIN SUPERUSER PASSWORD 'password';",,,,,,,,,
"psql.bin","ALTER ROLE",""

2017-07-16 13:00:28.469 EDT,"enterprisedb","edb",3356,"[local]",

596b9b7e.d1c,5,"idle",2017-07-16 12:59:42 EDT,6/9,0,AUDIT,00000,
"statement: CREATE DATABASE auditdb;",,,,,,,,,"psql.bin","CREATE DATABASE",""

2017-07-16 13:00:28.469 EDT,"enterprisedb","edb",3356,"[local]",

596b9b7e.d1c,6,"idle",2017-07-16 12:59:42 EDT,6/10,0,AUDIT,00000,
"statement: ALTER DATABASE auditdb OWNER TO adminuser;",,,,,,,,,"psql.bin","ALTER DATABASE",""

2017-07-16 13:01:13.735 EDT,"adminuser","auditdb",3377,"[local]",

596b9bd9.d31,1,"authentication",2017-07-16 13:01:13 EDT,4/15,0,AUDIT,00000,
"connection authorized: user=adminuser database=auditdb",,,,,,,,,"","",""

2017-07-16 13:01:13.735 EDT,"adminuser","auditdb",3377,"[local]",

596b9bd9.d31,2,"idle",2017-07-16 13:01:13 EDT,4/17,0,AUDIT,00000,
"statement: CREATE SCHEMA edb;",,,,,,,,,"psql.bin","CREATE SCHEMA",""

2017-07-16 13:01:13.735 EDT,"adminuser","auditdb",3377,"[local]",

596b9bd9.d31,3,"idle",2017-07-16 13:01:13 EDT,4/19,0,AUDIT,00000,
"statement: CREATE TABLE department (
deptno NUMBER(2) NOT NULL CONSTRAINT dept_pk PRIMARY KEY,
dname VARCHAR2(14) CONSTRAINT dept_dname_uq UNIQUE,
loc VARCHAR2(13)
);",,,,,,,,,"psql.bin","CREATE TABLE",""

2017-07-16 13:01:13.735 EDT,"adminuser","auditdb",3377,"[local]",

596b9bd9.d31,4,"idle",2017-07-16 13:01:13 EDT,4/21,0,AUDIT,00000,
"statement: CREATE TABLE dept (
deptno NUMBER(2) NOT NULL CONSTRAINT dept_pk PRIMARY KEY,
dname VARCHAR2(14) CONSTRAINT dept_dname_uq UNIQUE,
loc VARCHAR2(13)
);",,,,,,,,,"psql.bin","CREATE TABLE",""

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 132

 EDB Postgres Advanced Server Guide

The CREATE and ALTER statements for the adminuser role and auditdb database are
audited. The error for the ALTER ROLE adminuser statement is also logged since error
is included in the edb_audit_statement parameter.

Similarly, the CREATE statements for schema edb and tables department and dept are
audited.

Note that the DROP TABLE department statement is not in the audit log since there is
no edb_audit_statement setting that would result in the auditing of successfully
processed DROP statements such as ddl, all, or drop.

Example 2

The following is an example where edb_audit_connect and

edb_audit_statement are set with the following non-default values:

edb_audit_connect = 'all'
edb_audit_statement = create view,create materialized view,create sequence,grant'

Thus, only SQL statements invoked by the CREATE VIEW , CREATE MATERIALIZED
VIEW, CREATE SEQUENCE and GRANT commands are audited.

The database session that occurs is the following:

$ psql auditdb adminuser
Password for user adminuser:
psql.bin (10.0.1)
Type "help" for help.

auditdb=# SHOW edb_audit_connect;

edb_audit_connect
-------------------
all
(1 row)

auditdb=# SHOW edb_audit_statement;

edb_audit_statement
------------------------------------------------------------
create view,create materialized view,create sequence,grant
(1 row)

auditdb=# SET search_path TO edb;

SET
auditdb=# CREATE TABLE emp (
auditdb(# empno NUMBER(4) NOT NULL CONSTRAINT emp_pk PRIMARY KEY,
auditdb(# ename VARCHAR2(10),
auditdb(# job VARCHAR2(9),
auditdb(# mgr NUMBER(4),
auditdb(# hiredate DATE,
auditdb(# sal NUMBER(7,2) CONSTRAINT emp_sal_ck CHECK (sal > 0),
auditdb(# comm NUMBER(7,2),
auditdb(# deptno NUMBER(2) CONSTRAINT emp_ref_dept_fk
auditdb(# REFERENCES dept(deptno)
auditdb(# );
CREATE TABLE
auditdb=# CREATE VIEW salesemp AS
auditdb-# SELECT empno, ename, hiredate, sal, comm FROM emp WHERE job = 'SALESMAN';
CREATE VIEW
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 133
 EDB Postgres Advanced Server Guide
auditdb=# CREATE MATERIALIZED VIEW managers AS
auditdb-# SELECT empno, ename, hiredate, sal, comm FROM emp WHERE job = 'MANAGER';
SELECT 0
auditdb=# CREATE SEQUENCE next_empno START WITH 8000 INCREMENT BY 1;
CREATE SEQUENCE
auditdb=# GRANT ALL ON dept TO PUBLIC;
GRANT
auditdb=# GRANT ALL ON emp TO PUBLIC;
GRANT

The resulting audit log file contains the following.

Each audit log entry has been split and displayed across multiple lines, and a blank line
has been inserted between the audit log entries for more clarity in the appearance of the
results.
2017-07-16 13:20:09.836 EDT,"adminuser","auditdb",4143,"[local]",
596ba049.102f,1,"authentication",2017-07-16 13:20:09 EDT,4/10,0,AUDIT,00000,
"connection authorized: user=adminuser database=auditdb",,,,,,,,,"","",""

2017-07-16 13:20:09.836 EDT,"adminuser","auditdb",4143,"[local]",

596ba049.102f,2,"idle",2017-07-16 13:20:09 EDT,4/16,0,AUDIT,00000,
"statement: CREATE VIEW salesemp AS
SELECT empno, ename, hiredate, sal, comm FROM emp WHERE job =
'SALESMAN';",,,,,,,,,"psql.bin","CREATE VIEW",""

2017-07-16 13:20:09.836 EDT,"adminuser","auditdb",4143,"[local]",

596ba049.102f,3,"idle",2017-07-16 13:20:09 EDT,4/17,0,AUDIT,00000,
"statement: CREATE MATERIALIZED VIEW managers AS
SELECT empno, ename, hiredate, sal, comm FROM emp WHERE job =
'MANAGER';",,,,,,,,,"psql.bin","CREATE MATERIALIZED VIEW",""

2017-07-16 13:20:09.836 EDT,"adminuser","auditdb",4143,"[local]",

596ba049.102f,4,"idle",2017-07-16 13:20:09 EDT,4/18,0,AUDIT,00000,
"statement: CREATE SEQUENCE next_empno START WITH 8000 INCREMENT BY
1;",,,,,,,,,"psql.bin","CREATE SEQUENCE",""

2017-07-16 13:20:09.836 EDT,"adminuser","auditdb",4143,"[local]",

596ba049.102f,5,"idle",2017-07-16 13:20:09 EDT,4/19,0,AUDIT,00000,
"statement: GRANT ALL ON dept TO PUBLIC;",,,,,,,,,"psql.bin","GRANT",""

2017-07-16 13:20:09.836 EDT,"adminuser","auditdb",4143,"[local]",

596ba049.102f,6,"idle",2017-07-16 13:20:09 EDT,4/20,0,AUDIT,00000,
"statement: GRANT ALL ON emp TO PUBLIC;",,,,,,,,,"psql.bin","GRANT",""

The CREATE VIEW and CREATE MATERIALIZED VIEW statements are audited. Note
that the prior CREATE TABLE emp statement is not audited since none of the values
create, create table, ddl, nor all are included in the edb_audit_statement
parameter.

The CREATE SEQUENCE and GRANT statements are audited since those values are
included in the edb_audit_statement parameter.

3.5.2.2 Data Manipulation Language Statements

This section describes the values that can be included in the edb_audit_statement
parameter to audit DML statements.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 134

 EDB Postgres Advanced Server Guide

The following general rules apply:

If the edb_audit_statement parameter includes either dml or all, then all

DML statements are audited.
If the edb_audit_statement is set to none, then no DML statements are
audited.
Specific types of DML statements can be chosen for auditing by including a
combination of values within the edb_audit_statement parameter.

Use the following syntax to specify an edb_audit_statement parameter value for

SQL INSERT, UPDATE, DELETE, or TRUNCATE statements:

{ insert | update | delete | truncate }

Example

The following is an example where edb_audit_connect and

edb_audit_statement are set with the following non-default values:

edb_audit_connect = 'all'
edb_audit_statement = 'UPDATE, DELETE, error'

Thus, only SQL statements invoked by the UPDATE and DELETE commands are audited.
All errors are also included in the audit log (even errors not related to the UPDATE and
DELETE commands).

The database session that occurs is the following:

$ psql auditdb adminuser
Password for user adminuser:
psql.bin (10.0.1)
Type "help" for help.

auditdb=# SHOW edb_audit_connect;

edb_audit_connect
-------------------
all
(1 row)

auditdb=# SHOW edb_audit_statement;

edb_audit_statement
-----------------------
UPDATE, DELETE, error
(1 row)

auditdb=# SET search_path TO edb;

SET
auditdb=# INSERT INTO dept VALUES (10,'ACCOUNTING','NEW YORK');
INSERT 0 1
auditdb=# INSERT INTO dept VALUES (20,'RESEARCH','DALLAS');
INSERT 0 1
auditdb=# INSERT INTO dept VALUES (30,'SALES','CHICAGO');
INSERT 0 1
auditdb=# INSERT INTO dept VALUES (40,'OPERATIONS','BOSTON');
INSERT 0 1

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 135

 EDB Postgres Advanced Server Guide
auditdb=# INSERT INTO emp VALUES (7369,'SMITH','CLERK',7902,'17-DEC-80',800,NULL,20);
INSERT 0 1
auditdb=# INSERT INTO emp VALUES (7499,'ALLEN','SALESMAN',7698,'20-FEB-81',1600,300,30);
INSERT 0 1
auditdb=# INSERT INTO emp VALUES (7521,'WARD','SALESMAN',7698,'22-FEB-81',1250,500,30);
INSERT 0 1
.
.
.
auditdb=# INSERT INTO emp VALUES (7934,'MILLER','CLERK',7782,'23-JAN-82',1300,NULL,10);
INSERT 0 1
auditdb=# UPDATE dept SET loc = 'BEDFORD' WHERE deptno = 40;
UPDATE 1
auditdb=# SELECT * FROM dept;
deptno | dname | loc
--------+------------+----------
10 | ACCOUNTING | NEW YORK
20 | RESEARCH | DALLAS
30 | SALES | CHICAGO
40 | OPERATIONS | BEDFORD
(4 rows)

auditdb=# DELETE FROM emp WHERE deptno = 10;

DELETE 3
auditdb=# TRUNCATE employee;
ERROR: relation "employee" does not exist
auditdb=# TRUNCATE emp;
TRUNCATE TABLE
auditdb=# \q

The resulting audit log file contains the following.

Each audit log entry has been split and displayed across multiple lines, and a blank line
has been inserted between the audit log entries for more clarity in the appearance of the
results.
2017-07-16 13:43:26.638 EDT,"adminuser","auditdb",4574,"[local]",
596ba5be.11de,1,"authentication",2017-07-16 13:43:26 EDT,4/11,0,AUDIT,00000,
"connection authorized: user=adminuser database=auditdb",,,,,,,,,"","",""

2017-07-16 13:43:26.638 EDT,"adminuser","auditdb",4574,"[local]",

596ba5be.11de,2,"idle",2017-07-16 13:43:26 EDT,4/34,0,AUDIT,00000,
"statement: UPDATE dept SET loc = 'BEDFORD' WHERE deptno = 40;",,,,,,,,,"psql.bin","UPDATE",""

2017-07-16 13:43:26.638 EDT,"adminuser","auditdb",4574,"[local]",

596ba5be.11de,3,"idle",2017-07-16 13:43:26 EDT,4/36,0,AUDIT,00000,
"statement: DELETE FROM emp WHERE deptno = 10;",,,,,,,,,"psql.bin","DELETE",""

2017-07-16 13:45:46.999 EDT,"adminuser","auditdb",4574,"[local]",

596ba5be.11de,4,"TRUNCATE TABLE",2017-07-16 13:43:26 EDT,4/37,0,ERROR,42P01,
"relation ""employee"" does not exist",,,,,,"TRUNCATE employee;",,,"psql.bin","",""

2017-07-16 13:46:26.362 EDT,,,4491,,596ba59c.118b,1,,2017-07-16 13:42:52 EDT,,0,LOG,00000,

"database system is shut down",,,,,,,,,"","",""

The UPDATE dept and DELETE FROM emp statements are audited. Note that all of the
prior INSERT statements are not audited since none of the values insert, dml, nor all
are included in the edb_audit_statement parameter.

The SELECT * FROM dept statement is not audited as well since neither select nor
all is included in the edb_audit_statement parameter.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 136

 EDB Postgres Advanced Server Guide

Since error is specified in the edb_audit_statement parameter, but not the

truncate value, the error on the TRUNCATE employee statement is logged in the audit
file, but not the successful TRUNCATE emp statement.

3.5.3 Enabling Audit Logging

The following steps describe how to configure Advanced Server to log all connections,
disconnections, DDL statements, DCL statements, DML statements, and any statements
resulting in an error.

1. Enable auditing by the setting the edb_audit parameter to xml or csv.

2. Set the file rotation day when the new file will be created by setting the
parameter edb_audit_rotation_day to the desired value.
3. To audit all connections, set the parameter, edb_audit_connect, to all.
4. To audit all disconnections, set the parameter, edb_audit_disconnect, to
all.
5. To audit DDL, DCL, DML and other statements, set the parameter,
edb_audit_statement according to the instructions in Section 3.5.2.

The setting of the edb_audit_statement parameter in the configuration file affects

the entire database cluster.

The type of statements that are audited as controlled by the edb_audit_statement

parameter can be further refined according to the database in use as well as the database
role running the session:

The edb_audit_statement parameter can be set as an attribute of a specified

database with the ALTER DATABASE dbname SET edb_audit_statement
command. This setting overrides the edb_audit_statement parameter in the
configuration file for statements executed when connected to database dbname.
The edb_audit_statement parameter can be set as an attribute of a specified
role with the ALTER ROLE rolename SET edb_audit_statement
command. This setting overrides the edb_audit_statement parameter in the
configuration file as well as any setting assigned to the database by the previously
described ALTER DATABASE command when the specified role is running the
current session.
The edb_audit_statement parameter can be set as an attribute of a specified
role when using a specified database with the ALTER ROLE rolename IN
DATABASE dbname SET edb_audit_statement command. This setting
overrides the edb_audit_statement parameter in the configuration file as well
as any setting assigned to the database by the previously described ALTER
DATABASE command as well as any setting assigned to the role with the ALTER
ROLE command without the IN DATABASE clause as previously described.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 137

 EDB Postgres Advanced Server Guide

The following are examples of this technique.

The database cluster is established with edb_audit_statement set to all as shown in

its postgresql.conf file:
edb_audit_statement = 'all' # Statement type to be audited:
# none, dml, insert, update, delete, truncate,
# select, error, rollback, ddl, create, drop,
# alter, grant, revoke, all

A database and role are established with the following settings for the
edb_audit_statement parameter:

Database auditdb with ddl, insert, update, and delete

Role admin with select and truncate
Role admin in database auditdb with create table, insert, and update

Creation and alteration of the database and role are shown by the following:
$ psql edb enterprisedb
Password for user enterprisedb:
psql.bin (10.0.1)
Type "help" for help.

edb=# SHOW edb_audit_statement;

edb_audit_statement
---------------------
all
(1 row)

edb=# CREATE DATABASE auditdb;

CREATE DATABASE
edb=# ALTER DATABASE auditdb SET edb_audit_statement TO 'ddl, insert, update, delete';
ALTER DATABASE
edb=# CREATE ROLE admin WITH LOGIN SUPERUSER PASSWORD 'password';
CREATE ROLE
edb=# ALTER ROLE admin SET edb_audit_statement TO 'select, truncate';
ALTER ROLE
edb=# ALTER ROLE admin IN DATABASE auditdb SET edb_audit_statement TO 'create table, insert,
update';
ALTER ROLE

The following demonstrates the changes made and the resulting audit log file for three
cases.

Case 1: Changes made in database auditdb by role enterprisedb. Only ddl,

insert, update, and delete statements are audited:

$ psql auditdb enterprisedb

Password for user enterprisedb:
psql.bin (10.0.1)
Type "help" for help.

auditdb=# SHOW edb_audit_statement;

edb_audit_statement
-----------------------------
ddl, insert, update, delete
(1 row)
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 138
 EDB Postgres Advanced Server Guide

auditdb=# CREATE TABLE audit_tbl (f1 INTEGER PRIMARY KEY, f2 TEXT);

CREATE TABLE
auditdb=# INSERT INTO audit_tbl VALUES (1, 'Row 1');
INSERT 0 1
auditdb=# UPDATE audit_tbl SET f2 = 'Row A' WHERE f1 = 1;
UPDATE 1
auditdb=# SELECT * FROM audit_tbl; <== Should not be audited
f1 | f2
----+-------
1 | Row A
(1 row)

auditdb=# TRUNCATE audit_tbl; <== Should not be audited

TRUNCATE TABLE

The following audit log file shows entries only for the CREATE TABLE, INSERT INTO
audit_tbl, and UPDATE audit_tbl statements. The SELECT * FROM audit_tbl
and TRUNCATE audit_tbl statements were not audited.

Each audit log entry has been split and displayed across multiple lines, and a blank line
has been inserted between the audit log entries for more clarity in the appearance of the
results.
2017-07-13 15:26:17.426 EDT,"enterprisedb","auditdb",4024,"[local]",
5967c947.fb8,1,"idle",2017-07-13 15:25:59 EDT,7/4,0,AUDIT,00000,
"statement: CREATE TABLE audit_tbl (f1 INTEGER PRIMARY KEY, f2 TEXT);",,,,,,,,,
"psql.bin","CREATE TABLE",""

2017-07-13 15:26:17.426 EDT,"enterprisedb","auditdb",4024,"[local]",

5967c947.fb8,2,"idle",2017-07-13 15:25:59 EDT,7/5,0,AUDIT,00000,
"statement: INSERT INTO audit_tbl VALUES (1, 'Row 1');",,,,,,,,,"psql.bin","INSERT",""

2017-07-13 15:26:17.426 EDT,"enterprisedb","auditdb",4024,"[local]",

5967c947.fb8,3,"idle",2017-07-13 15:25:59 EDT,7/6,0,AUDIT,00000,
"statement: UPDATE audit_tbl SET f2 = 'Row A' WHERE f1 = 1;",,,,,,,,,"psql.bin","UPDATE",""

Case 2: Changes made in database edb by role admin. Only select and truncate
statements are audited:
$ psql edb admin
Password for user admin:
psql.bin (10.0.1)
Type "help" for help.

edb=# SHOW edb_audit_statement;

edb_audit_statement
---------------------
select, truncate
(1 row)

edb=# CREATE TABLE edb_tbl (f1 INTEGER PRIMARY KEY, f2 TEXT) <== Should not be audited
CREATE TABLE
edb=# INSERT INTO edb_tbl VALUES (1, 'Row 1'); <== Should not be audited
INSERT 0 1
edb=# SELECT * FROM edb_tbl;
f1 | f2
----+-------
1 | Row 1
(1 row)

edb=# TRUNCATE edb_tbl;

TRUNCATE TABLE

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 139

 EDB Postgres Advanced Server Guide

Continuation of the audit log file now appears as follows. The last two entries
representing the second case show only the SELECT * FROM edb_tbl and TRUNCATE
edb_tbl statements. The CREATE TABLE edb_tbl and INSERT INTO edb_tbl
statements were not audited.
2017-07-13 15:26:17.426 EDT,"enterprisedb","auditdb",4024,"[local]",
5967c947.fb8,1,"idle",2017-07-13 15:25:59 EDT,7/4,0,AUDIT,00000,
"statement: CREATE TABLE audit_tbl (f1 INTEGER PRIMARY KEY, f2 TEXT);",,,,,,,,,
"psql.bin","CREATE TABLE",""

2017-07-13 15:26:17.426 EDT,"enterprisedb","auditdb",4024,"[local]",

5967c947.fb8,2,"idle",2017-07-13 15:25:59 EDT,7/5,0,AUDIT,00000,
"statement: INSERT INTO audit_tbl VALUES (1, 'Row 1');",,,,,,,,,"psql.bin","INSERT",""

2017-07-13 15:26:17.426 EDT,"enterprisedb","auditdb",4024,"[local]",

5967c947.fb8,3,"idle",2017-07-13 15:25:59 EDT,7/6,0,AUDIT,00000,
"statement: UPDATE audit_tbl SET f2 = 'Row A' WHERE f1 = 1;",,,,,,,,,"psql.bin","UPDATE",""

2017-07-13 15:29:45.616 EDT,"admin","edb",4047,"[local]",

5967ca05.fcf,1,"idle",2017-07-13 15:29:09 EDT,4/33,0,AUDIT,00000,
"statement: SELECT * FROM edb_tbl;",,,,,,,,,"psql.bin","SELECT",""

2017-07-13 15:29:45.616 EDT,"admin","edb",4047,"[local]",

5967ca05.fcf,2,"idle",2017-07-13 15:29:09 EDT,4/34,0,AUDIT,00000,
"statement: TRUNCATE edb_tbl;",,,,,,,,,"psql.bin","TRUNCATE TABLE",""

Case 3: Changes made in database auditdb by role admin. Only create table,
insert, and update statements are audited:

$ psql auditdb admin

Password for user admin:
psql.bin (10.0.1)
Type "help" for help.

auditdb=# SHOW edb_audit_statement;

edb_audit_statement
------------------------------
create table, insert, update
(1 row)

auditdb=# CREATE TABLE audit_tbl_2 (f1 INTEGER PRIMARY KEY, f2 TEXT);

CREATE TABLE
auditdb=# INSERT INTO audit_tbl_2 VALUES (1, 'Row 1');
INSERT 0 1
auditdb=# SELECT * FROM audit_tbl_2; <== Should not be audited
f1 | f2
----+-------
1 | Row 1
(1 row)

auditdb=# TRUNCATE audit_tbl_2; <== Should not be audited

TRUNCATE TABLE

Continuation of the audit log file now appears as follows. The next to last two entries
representing the third case show only CREATE TABLE audit_tbl_2 and INSERT
INTO audit_tbl_2 statements. The SELECT * FROM audit_tbl_2 and TRUNCATE
audit_tbl_2 statements were not audited.

2017-07-13 15:26:17.426 EDT,"enterprisedb","auditdb",4024,"[local]",

5967c947.fb8,1,"idle",2017-07-13 15:25:59 EDT,7/4,0,AUDIT,00000,
"statement: CREATE TABLE audit_tbl (f1 INTEGER PRIMARY KEY, f2 TEXT);",,,,,,,,,
"psql.bin","CREATE TABLE",""
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 140
 EDB Postgres Advanced Server Guide

2017-07-13 15:26:17.426 EDT,"enterprisedb","auditdb",4024,"[local]",

5967c947.fb8,2,"idle",2017-07-13 15:25:59 EDT,7/5,0,AUDIT,00000,
"statement: INSERT INTO audit_tbl VALUES (1, 'Row 1');",,,,,,,,,"psql.bin","INSERT",""

2017-07-13 15:26:17.426 EDT,"enterprisedb","auditdb",4024,"[local]",

5967c947.fb8,3,"idle",2017-07-13 15:25:59 EDT,7/6,0,AUDIT,00000,
"statement: UPDATE audit_tbl SET f2 = 'Row A' WHERE f1 = 1;",,,,,,,,,"psql.bin","UPDATE",""

2017-07-13 15:29:45.616 EDT,"admin","edb",4047,"[local]",

5967ca05.fcf,1,"idle",2017-07-13 15:29:09 EDT,4/33,0,AUDIT,00000,
"statement: SELECT * FROM edb_tbl;",,,,,,,,,"psql.bin","SELECT",""

2017-07-13 15:29:45.616 EDT,"admin","edb",4047,"[local]",

5967ca05.fcf,2,"idle",2017-07-13 15:29:09 EDT,4/34,0,AUDIT,00000,
"statement: TRUNCATE edb_tbl;",,,,,,,,,"psql.bin","TRUNCATE TABLE",""

2017-07-13 15:35:45.309 EDT,"admin","auditdb",4085,"[local]",

5967cb81.ff5,1,"idle",2017-07-13 15:35:29 EDT,4/72,0,AUDIT,00000,
"statement: CREATE TABLE audit_tbl_2 (f1 INTEGER PRIMARY KEY, f2 TEXT);",,,,,,,,,
"psql.bin","CREATE TABLE",""

2017-07-13 15:35:45.309 EDT,"admin","auditdb",4085,"[local]",

5967cb81.ff5,2,"idle",2017-07-13 15:35:29 EDT,4/73,0,AUDIT,00000,
"statement: INSERT INTO audit_tbl_2 VALUES (1, 'Row 1');",,,,,,,,,"psql.bin","INSERT",""

2017-07-13 15:38:42.028 EDT,,,3942,,5967c934.f66,1,,2017-07-13 15:25:40

EDT,,0,LOG,00000,"database system is shut down",,,,,,,,,"","",""

3.5.4 Audit Log File

The audit log file can be generated in either CSV or XML format depending upon the
setting of the edb_audit configuration parameter. The XML format contains less
information than the CSV format.

The information in the audit log is based on the logging performed by PostgreSQL as
described in Section 19.8.4 Using CSV-Format Log Output within Section 19.8 Error
Reporting and Logging in the PostgreSQL core documentation, available at:

https://www.postgresql.org/docs/10/static/runtime-config-logging.html

The following table lists the fields in the order they appear in the CSV audit log format.
The table contains the following information:

Field. Name of the field as shown in the sample table definition in the
PostgreSQL documentation as previously referenced.
XML Element/Attribute. For the XML format, name of the XML element and
its attribute (if used), referencing the value. Note: n/a indicates that there is no
XML representation for this field.
Data Type. Data type of the field as given by the PostgreSQL sample table
definition.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 141

 EDB Postgres Advanced Server Guide

Description. Description of the field. For certain fields, no output is generated in

the audit log as those fields are not supported by auditing. Those fields are
indicated by Not supported.

The fields with the Description of Not supported appear as consecutive commas
(,,) in the CSV format.

Table 3-3 - Audit Log Fields

Field XML Element/Attribute Data Type Description
log_time event/time timestamp Log date/time of the
with time statement.
zone
user_name event/user text Database user who executed
the statement.
database_name event/database text Database in which the
statement was executed.
process_id event/process_id integer Operating system process ID
in which the statement was
executed.
connection_from event/remote_host text Host and port location from
where the statement was
executed.
session_id event/session_id text Session ID in which the
statement was executed.
session_line_num n/a bigint Order of the statement within
the session.
process_status n/a text Processing status.
session_start_time n/a timestamp Date/time when the session
with time was started.
zone
virtual_transaction_id n/a text Virtual transaction ID of the
statement.
transaction_id event/transaction_id bigint Regular transaction ID of the
statement.
error_severity message text Statement severity. Values
are AUDIT for audited
statements and ERROR for any
resulting error messages.
sql_state_code n/a text SQL state code returned for
the statement.
message message text The SQL statement that was
attempted for execution.
detail n/a text Error message detail. (Not
supported)
hint n/a text Hint (Not supported)
internal_query n/a text Internal query that led to the
error, if any. (Not supported)
internal_query_pos n/a integer Character count of the error
position, therein. (Not
supported)
context n/a text Error context. (Not
supported)
query n/a text User query that led to the
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 142
 EDB Postgres Advanced Server Guide

Field XML Element/Attribute Data Type Description

error. (For errors only)
query_pos n/a integer Character count of the error
position, therein. (For errors
only)
location n/a text Location of the error in the
PostgreSQL source code.
(Not supported)
application_name n/a text Name of the application from
which the statement was
executed. (for example,
psql.bin).
command_tag event/command_tag text SQL command of the
statement.
audit_tag event/audit_tag text Value specified by the
audit_tag parameter in the
configuration file.

The following examples are generated in the CSV and XML formats.

The non-default audit settings in the postgresql.conf file are as follows:

edb_audit = 'csv'
edb_audit_connect = 'all'
edb_audit_disconnect = 'all'
edb_audit_statement = 'ddl, dml, select, error'
edb_audit_tag = 'edbaudit'

The edb_audit parameter is changed to xml when generating the XML format.

The audited session is the following:

$ psql edb enterprisedb
Password for user enterprisedb:
psql.bin (10.0.1)
Type "help" for help.

edb=# CREATE SCHEMA edb;

CREATE SCHEMA
edb=# SET search_path TO edb;
SET
edb=# CREATE TABLE dept (
edb(# deptno NUMBER(2) NOT NULL CONSTRAINT dept_pk PRIMARY KEY,
edb(# dname VARCHAR2(14) CONSTRAINT dept_dname_uq UNIQUE,
edb(# loc VARCHAR2(13)
edb(# );
CREATE TABLE
edb=# INSERT INTO dept VALUES (10,'ACCOUNTING','NEW YORK');
INSERT 0 1
edb=# UPDATE department SET loc = 'BOSTON' WHERE deptno = 10;
ERROR: relation "department" does not exist
LINE 1: UPDATE department SET loc = 'BOSTON' WHERE deptno = 10;
^
edb=# UPDATE dept SET loc = 'BOSTON' WHERE deptno = 10;
UPDATE 1
edb=# SELECT * FROM dept;
deptno | dname | loc
--------+------------+--------
10 | ACCOUNTING | BOSTON
(1 row)
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 143
 EDB Postgres Advanced Server Guide

edb=# \q

CSV Audit Log File

The following is the CSV format of the audit log file.

Each audit log entry has been split and displayed across multiple lines, and a blank line
has been inserted between the audit log entries for more clarity in the appearance of the
results.
2017-07-17 13:28:44.235 EDT,"enterprisedb","edb",4068,"[local]",
596cf3cc.fe4,1,"authentication",2017-07-17 13:28:44 EDT,6/2,0,AUDIT,00000,
"connection authorized: user=enterprisedb database=edb",,,,,,,,,"","","edbaudit"

2017-07-17 13:28:44.235 EDT,"enterprisedb","edb",4068,"[local]",

596cf3cc.fe4,2,"idle",2017-07-17 13:28:44 EDT,6/4,0,AUDIT,00000,
"statement: CREATE SCHEMA edb;",,,,,,,,,"psql.bin","CREATE SCHEMA","edbaudit"

2017-07-17 13:28:44.235 EDT,"enterprisedb","edb",4068,"[local]",

596cf3cc.fe4,3,"idle",2017-07-17 13:28:44 EDT,6/6,0,AUDIT,00000,
"statement: CREATE TABLE dept (
deptno NUMBER(2) NOT NULL CONSTRAINT dept_pk PRIMARY KEY,
dname VARCHAR2(14) CONSTRAINT dept_dname_uq UNIQUE,
loc VARCHAR2(13)
);",,,,,,,,,"psql.bin","CREATE TABLE","edbaudit"

2017-07-17 13:28:44.235 EDT,"enterprisedb","edb",4068,"[local]",

596cf3cc.fe4,4,"idle",2017-07-17 13:28:44 EDT,6/7,0,AUDIT,00000,
"statement: INSERT INTO dept VALUES (10,'ACCOUNTING','NEW YORK');",,,,,,,,,
"psql.bin","INSERT","edbaudit"

2017-07-17 13:28:44.235 EDT,"enterprisedb","edb",4068,"[local]",

596cf3cc.fe4,5,"idle",2017-07-17 13:28:44 EDT,6/8,0,AUDIT,00000,
"statement: UPDATE department SET loc = 'BOSTON' WHERE deptno = 10;",,,,,,,,,
"psql.bin","UPDATE","edbaudit"

2017-07-17 13:29:59.833 EDT,"enterprisedb","edb",4068,"[local]",

596cf3cc.fe4,6,"UPDATE",2017-07-17 13:28:44 EDT,6/8,0,ERROR,42P01,
"relation ""department"" does not exist",,,,,,
"UPDATE department SET loc = 'BOSTON' WHERE deptno = 10;",8,,"psql.bin","","edbaudit"

2017-07-17 13:29:59.833 EDT,"enterprisedb","edb",4068,"[local]",

596cf3cc.fe4,7,"idle",2017-07-17 13:28:44 EDT,6/9,0,AUDIT,00000,
"statement: UPDATE dept SET loc = 'BOSTON' WHERE deptno = 10;",,,,,,,,,
"psql.bin","UPDATE","edbaudit"

2017-07-17 13:29:59.833 EDT,"enterprisedb","edb",4068,"[local]",

596cf3cc.fe4,8,"idle",2017-07-17 13:28:44 EDT,6/10,0,AUDIT,00000,
"statement: SELECT * FROM dept;",,,,,,,,,"psql.bin","SELECT","edbaudit"

2017-07-17 13:29:59.833 EDT,"enterprisedb","edb",4068,"[local]",

596cf3cc.fe4,9,"idle",2017-07-17 13:28:44 EDT,,0,AUDIT,00000,
"disconnection: session time: 0:02:01.511 user=enterprisedb database=edb
host=[local]",,,,,,,,,"psql.bin","SELECT","edbaudit"

2017-07-17 13:30:53.617 EDT,,,3987,,596cf3b3.f93,1,,2017-07-17 13:28:19 EDT,,0,LOG,00000,

"database system is shut down",,,,,,,,,"","","edbaudit"

XML Audit Log File

The following is the XML format of the audit log file. The output has been formatted for
more clarity in the appearance in the example.
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 144
 EDB Postgres Advanced Server Guide
<event user="enterprisedb" database="edb" remote_host="[local]"
session_id="596cf5b7.12a8" process_id="4776" time="2017-07-17 13:36:55 EDT"
transaction_id="0" type="connect" audit_tag="edbaudit">
<message>AUDIT: connection authorized: user=enterprisedb database=edb</message>
</event>
<event user="enterprisedb" database="edb" remote_host="[local]"
session_id="596cf5b7.12a8" process_id="4776" time="2017-07-17 13:37:02 EDT"
transaction_id="0" type="create" command_tag="CREATE SCHEMA" audit_tag="edbaudit">
<message>AUDIT: statement: CREATE SCHEMA edb;</message>
</event>
<event user="enterprisedb" database="edb" remote_host="[local]"
session_id="596cf5b7.12a8" process_id="4776" time="2017-07-17 13:37:19 EDT"
transaction_id="0" type="create" command_tag="CREATE TABLE" audit_tag="edbaudit">
<message>AUDIT: statement: CREATE TABLE dept (
deptno NUMBER(2) NOT NULL CONSTRAINT dept_pk PRIMARY KEY,
dname VARCHAR2(14) CONSTRAINT dept_dname_uq UNIQUE,
loc VARCHAR2(13));
</message>
</event>
<event user="enterprisedb" database="edb" remote_host="[local]"
session_id="596cf5b7.12a8" process_id="4776" time="2017-07-17 13:37:29 EDT"
transaction_id="0" type="insert" command_tag="INSERT" audit_tag="edbaudit">
<message>AUDIT: statement: INSERT INTO dept VALUES
(10,&apos;ACCOUNTING&apos;,&apos;NEW YORK&apos;);
</message>
</event>
<event user="enterprisedb" database="edb" remote_host="[local]"
session_id="596cf5b7.12a8" process_id="4776" time="2017-07-17 13:37:40 EDT"
transaction_id="0" type="update" command_tag="UPDATE" audit_tag="edbaudit">
<message>AUDIT: statement: UPDATE department SET
loc = &apos;BOSTON&apos; WHERE deptno = 10;
</message>
</event>
<event user="enterprisedb" database="edb" remote_host="[local]"
session_id="596cf5b7.12a8" process_id="4776" time="2017-07-17 13:37:40 EDT"
transaction_id="0" type="error" audit_tag="edbaudit">
<message>ERROR: relation &quot;department&quot; does not exist at character 8
</message>
</event>
<event user="enterprisedb" database="edb" remote_host="[local]"
session_id="596cf5b7.12a8" process_id="4776" time="2017-07-17 13:37:51 EDT"
transaction_id="0" type="update" command_tag="UPDATE" audit_tag="edbaudit">
<message>AUDIT: statement: UPDATE dept SET loc = &apos;BOSTON&apos; WHERE deptno = 10;
</message>
</event>
<event user="enterprisedb" database="edb" remote_host="[local]"
session_id="596cf5b7.12a8" process_id="4776" time="2017-07-17 13:37:59 EDT"
transaction_id="0" type="select" command_tag="SELECT" audit_tag="edbaudit">
<message>AUDIT: statement: SELECT * FROM dept;</message>
</event>
<event user="enterprisedb" database="edb" remote_host="[local]"
session_id="596cf5b7.12a8" process_id="4776" time="2017-07-17 13:38:01 EDT"
transaction_id="0" type="disconnect" command_tag="SELECT" audit_tag="edbaudit">
<message>AUDIT: disconnection: session time: 0:01:05.814
user=enterprisedb database=edb host=[local]
</message>
</event>
<event process_id="4696" time="2017-07-17 13:38:08 EDT"
transaction_id="0" type="shutdown" audit_tag="edbaudit">
<message>LOG: database system is shut down</message>
</event>

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 145

 EDB Postgres Advanced Server Guide

3.5.5 Filtering Audit Logs Using Error Codes

Advanced Server includes an extension that you can use to exclude log file entries that
include a user-specified error code from the Advanced Server log files. To filter audit log
entries, you must first enable the extension by modifying the postgresql.conf file,
adding the following value to the values specified in the shared_preload_libraries
parameter:

$libdir/edb_filter_log

Then, use the edb_filter_log.errcodes parameter to specify any error codes you
wish to omit from the log files:

edb_filter_log.errcode = 'error_code'

Where error_code specifies one or more error codes that you wish to omit from the log
file. Provide multiple error codes in a comma-delimited list.

For example, if edb_filter_log is enabled, and edb_filter_log.errcode is set to

'23505,23502,22012', any log entries that return one of the following SQLSTATE
errors:

23505 (for violating a unique constraint)

23502 (for violating a not-null constraint)

22012 (for dividing by zero)

will be omitted from the log file.

For a complete list of the error codes supported by Advanced Server audit log filtering,
please see the core documentation at:

https://www.postgresql.org/docs/10/static/errcodes-appendix.html

3.5.6 Filtering Audit Logs Using Command Tags

Each entry in the log file except for those displaying an error message contains a
command tag, which is the SQL command executed for that particular log entry.

The command tag makes it possible to use subsequent tools to scan the log file to find
entries related to a particular SQL command.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 146

 EDB Postgres Advanced Server Guide

The following is an example in XML form. The output has been formatted for easier
appearance in the example.

The command tag is displayed as the command_tag attribute of the event element with
values CREATE ROLE, ALTER ROLE, and DROP ROLE in the example.
<event user="enterprisedb" database="edb" remote_host="[local]"
session_id="595e8537.10f1" process_id="4337" time="2017-07-06 14:45:18 EDT"
transaction_id="0" type="create"
command_tag="CREATE ROLE">
<message>AUDIT: statement: CREATE ROLE newuser WITH LOGIN;</message>
</event>
<event user="enterprisedb" database="edb" remote_host="[local]"
session_id="595e8537.10f1" process_id="4337" time="2017-07-06 14:45:31 EDT"
transaction_id="0" type="error">
<message>ERROR: unrecognized role option &quot;super&quot; at character 25
STATEMENT: ALTER ROLE newuser WITH SUPER USER;</message>
</event>
<event user="enterprisedb" database="edb" remote_host="[local]"
session_id="595e8537.10f1" process_id="4337" time="2017-07-06 14:45:38 EDT"
transaction_id="0" type="alter" command_tag="ALTER ROLE">
<message>AUDIT: statement: ALTER ROLE newuser WITH SUPERUSER;</message>
</event>
<event user="enterprisedb" database="edb" remote_host="[local]"
session_id="595e8537.10f1" process_id="4337" time="2017-07-06 14:45:46 EDT"
transaction_id="0" type="drop" command_tag="DROP ROLE">
<message>AUDIT: statement: DROP ROLE newuser;</message>
</event>

The following is the same example in CSV form. The command tag is the next to last
column of each entry. (The last column appears empty as "", which would be the value
provided by the edb_audit_tag parameter.)

Each audit log entry has been split and displayed across multiple lines, and a blank line
has been inserted between the audit log entries for more clarity in the appearance of the
results.
2017-07-06 14:47:22.294 EDT,"enterprisedb","edb",4720,"[local]",
595e85b2.1270,1,"idle",2017-07-06 14:47:14 EDT,6/4,0,AUDIT,00000,
"statement: CREATE ROLE newuser WITH LOGIN;",,,,,,,,,"psql.bin","CREATE ROLE",""

2017-07-06 14:47:29.694 EDT,"enterprisedb","edb",4720,"[local]",

595e85b2.1270,2,"idle",2017-07-06 14:47:14 EDT,6/5,0,ERROR,42601,
"unrecognized role option ""super""",,,,,,"ALTER ROLE newuser WITH SUPER USER;",25,,
"psql.bin","",""

2017-07-06 14:47:29.694 EDT,"enterprisedb","edb",4720,"[local]",

595e85b2.1270,3,"idle",2017-07-06 14:47:14 EDT,6/6,0,AUDIT,00000,
"statement: ALTER ROLE newuser WITH SUPERUSER;",,,,,,,,,"psql.bin","ALTER ROLE",""

2017-07-06 14:47:29.694 EDT,"enterprisedb","edb",4720,"[local]",

595e85b2.1270,4,"idle",2017-07-06 14:47:14 EDT,6/7,0,AUDIT,00000,
"statement: DROP ROLE newuser;",,,,,,,,,"psql.bin","DROP ROLE",""

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 147

 EDB Postgres Advanced Server Guide

3.6 Unicode Collation Algorithm

The Unicode Collation Algorithm (UCA) is a specification (Unicode Technical Report
#10) that defines a customizable method of collating and comparing Unicode data.
Collation means how data is sorted as with a SELECT ORDER BY clause. Comparison
is relevant for searches that use ranges with less than, greater than, or equal to operators.

Customizability is an important factor for various reasons such as the following.

Unicode supports a vast number of languages. Letters that may be common to

several languages may be expected to collate in different orders depending upon
the language.
Characters that appear with letters in certain languages such as accents or umlauts
have an impact on the expected collation depending upon the language.
In some languages, combinations of several consecutive characters should be
treated as a single character with regards to its collation sequence.
There may be certain preferences as to the collation of letters according to case.
For example, should the lowercase form of a letter collate before the uppercase
form of the same letter or vice versa.
There may be preferences as to whether punctuation marks such as underscore
characters, hyphens, or space characters should be considered in the collating
sequence or should they simply be ignored as if they did not exist in the string.

Given all of these variations with the vast number of languages supported by Unicode,
there is a necessity for a method to select the specific criteria for determining a collating
sequence. This is what the Unicode Collation Algorithm defines.

Note: In addition, another advantage for using ICU collations (the implementation of the
Unicode Collation Algorithm) is for performance. Sorting tasks, including B-tree index
creation, can complete in less than half the time it takes with a non-ICU collation. The
exact performance gain depends on your operating system version, the language of your
text data, and other factors.

The following sections provide a brief, simplified explanation of the Unified Collation
Algorithm concepts. As the algorithm and its usage are quite complex with numerous
variations, refer to the official documents cited in these sections for complete details.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 148

 EDB Postgres Advanced Server Guide

3.6.1 Basic Unicode Collation Algorithm Concepts

The official information for the Unicode Collation Algorithm is specified in Unicode
Technical Report #10, which can be found on The Unicode Consortium website at:

http://www.unicode.org/reports/tr10/

The ICU International Components for Unicode also provides much useful information.
An explanation of the collation concepts can be found on their website located at:

http://userguide.icu-project.org/collation/concepts

The basic concept behind the Unicode Collation Algorithm is the use of multilevel
comparison. This means that a number of levels are defined, which are listed as level 1
through level 5 in the following bullet points. Each level defines a type of comparison.
Strings are first compared using the primary level, also called level 1.

If the order can be determined based on the primary level, then the algorithm is done. If
the order cannot be determined based on the primary level, then the secondary level, level
2, is applied. If the order can be determined based on the secondary level, then the
algorithm is done, otherwise the tertiary level is applied, and so on. There is typically, a
final tie-breaking level to determine the order if it cannot be resolved by the prior levels.

Level 1 Primary Level for Base Characters. The order of basic characters
such as letters and digits determines the difference such as A < B.
Level 2 Secondary Level for Accents. If there are no primary level differences,
then the presence or absence of accents and other such characters determine the
order such as a < .
Level 3 Tertiary Level for Case. If there are no primary level or secondary
level differences, then a difference in case determines the order such as a < A.
Level 4 Quaternary Level for Punctuation. If there are no primary,
secondary, or tertiary level differences, then the presence or absence of white
space characters, control characters, and punctuation determine the order such as
-A < A.
Level 5 Identical Level for Tie-Breaking. If there are no primary, secondary,
tertiary, or quaternary level differences, then some other difference such as the
code point values determines the order.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 149

 EDB Postgres Advanced Server Guide

3.6.2 International Components for Unicode

The Unicode Collation Algorithm is implemented by open source software provided by

the International Components for Unicode (ICU). The software is a set of C/C++ and
Java libraries.

When Advanced Server is used to create a collation that invokes the ICU components to
produce the collation, the result is referred to as an ICU collation.

3.6.2.1 Locale Collations

When creating a collation for a locale, a predefined ICU short form name for the given
locale is typically provided.

An ICU short form is a method of specifying collation attributes, which are the
properties of a collation. Section 3.6.2.2 provides additional information on collation
attributes.

There are predefined ICU short forms for locales. The ICU short form for a locale
incorporates the collation attribute settings typically used for the given locale. This
simplifies the collation creation process by eliminating the need to specify the entire list
of collation attributes for that locale.

The system catalog pg_catalog.pg_icu_collate_names contains a list of the

names of the ICU short forms for locales. The ICU short form name is listed in column
icu_short_form.

edb=# SELECT icu_short_form, valid_locale FROM pg_icu_collate_names ORDER BY

valid_locale;
icu_short_form | valid_locale
----------------+--------------
LAF | af
LAR | ar
LAS | as
LAZ | az
LBE | be
LBG | bg
LBN | bn
LBS | bs
LBS_ZCYRL | bs_Cyrl
LROOT | ca
LROOT | chr
LCS | cs
LCY | cy
LDA | da
LROOT | de
LROOT | dz
LEE | ee
LEL | el

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 150

 EDB Postgres Advanced Server Guide

LROOT | en
LROOT | en_US
LEN_RUS_VPOSIX | en_US_POSIX
LEO | eo
LES | es
LET | et
LFA | fa
LFA_RAF | fa_AF
.
.
.

If needed, the default characteristics of an ICU short form for a given locale can be
overridden by specifying the collation attributes to override that property. This is
discussed in the next section.

3.6.2.2 Collation Attributes

When creating an ICU collation, the desired characteristics of the collation must be
specified. As discussed in Section 3.6.2.1, this can typically be done with an ICU short
form for the desired locale. However, if more specific information is required, the
specification of the collation properties can be done by using collation attributes.

Collation attributes define the rules of how characters are to be compared for determining
the collation sequence of text strings. As Unicode covers a vast set of languages in
numerous variations according to country, territory and culture, these collation attributes
are quite complex.

For the complete, precise meaning and usage of collation attributes, see Section 13
Collator Naming Scheme on the ICU International Components for Unicode website
at:

http://userguide.icu-project.org/collation/concepts

The following is a brief summary of the collation attributes and how they are specified
using the ICU short form method

Each collation attribute is represented by an uppercase letter, which are listed in the
following bullet points. The possible valid values for each attribute are given by codes
shown within the parentheses. Some codes have general meanings for all attributes. X
means to set the attribute off. O means to set the attribute on. D means to set the attribute
to its default value.

A Alternate (N, S, D). Handles treatment of variable characters such as white

spaces, punctuation marks, and symbols. When set to non-ignorable (N),
differences in variable characters are treated with the same importance as
differences in letters. When set to shifted (S), then differences in variable
characters are of minor importance (that is, the variable character is ignored when
comparing base characters).
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 151
 EDB Postgres Advanced Server Guide

C Case First (X, L, U, D). Controls whether a lowercase letter sorts before the
same uppercase letter (L), or the uppercase letter sorts before the same lowercase
letter (U). Off (X) is typically specified when lowercase first (L) is desired.
E Case Level (X, O, D). Set in combination with the Strength attribute, the
Case Level attribute is used when accents are to be ignored, but not case.
F French Collation (X, O, D). When set to on, secondary differences (presence
of accents) are sorted from the back of the string as done in the French Canadian
locale.
H Hiragana Quaternary (X, O, D). Introduces an additional level to
distinguish between the Hiragana and Katakana characters for compatibility with
the JIS X 4061 collation of Japanese character strings.
N Normalization Checking (X, O, D). Controls whether or not text is
thoroughly normalized for comparison. Normalization deals with the issue of
canonical equivalence of text whereby different code point sequences represent
the same character, which then present issues when sorting or comparing such
characters. Languages such as Arabic, ancient Greek, Hebrew, Hindi, Thai, or
Vietnamese should be used with Normalization Checking set to on.
S Strength (1, 2, 3, 4, I, D). Maximum collation level used for comparison.
Influences whether accents or case are taken into account when collating or
comparing strings. Each number represents a level. A setting of I represents
identical strength (that is, level 5).
T Variable Top (hexadecimal digits). Applicable only when the Alternate
attribute is not set to non-ignorable (N). The hexadecimal digits specify the
highest character sequence that is to be considered ignorable. For example, if
white space is to be ignorable, but visible variable characters are not to be
ignorable, then Variable Top set to 0020 would be specified along with the
Alternate attribute set to S and the Strength attribute set to 3. (The space character
is hexadecimal 0020. Other non-visible variable characters such as backspace,
tab, line feed, carriage return, etc. have values less than 0020. All visible
punctuation marks have values greater than 0020.)

A set of collation attributes and their values is represented by a text string consisting of
the collation attribute letter concatenated with the desired attribute value. Each
attribute/value pair is joined to the next pair with an underscore character as shown by the
following example.

AN_CX_EX_FX_HX_NO_S3

Collation attributes can be specified along with a locales ICU short form name to
override those default attribute settings of the locale.

The following is an example where the ICU short form named LROOT is modified with a
number of other collation attribute/value pairs.

AN_CX_EX_LROOT_NO_S3

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 152

 EDB Postgres Advanced Server Guide

In the preceding example, the Alternate attribute (A) is set to non-ignorable (N). The Case
First attribute (C) is set to off (X). The Case Level attribute (E) is set to off (X). The
Normalization attribute (N) is set to on (O). The Strength attribute (S) is set to the tertiary
level 3. LROOT is the ICU short form to which these other attributes are applying
modifications.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 153

 EDB Postgres Advanced Server Guide

3.6.3 Creating an ICU Collation

Creating an ICU collation can be done a number of different ways.

When creating a new database cluster with the initdb command, the --icu-
short-form option can be specified to define the ICU collation to be used by
default by all databases in the cluster.
When creating a new database with the CREATE DATABASE command, the
ICU_SHORT_FORM parameter can be specified to define the ICU collation to be
used by default in that database.
In an existing database, the CREATE COLLATION command can be used with the
ICU_SHORT_FORM parameter to define an ICU collation to be used under specific
circumstances such as when assigned with the COLLATE clause onto selected
columns of certain tables or when appended with the COLLATE clause onto an
expression such as ORDER BY expr COLLATE "collation_name".

The following describes the various ways of creating an ICU collation.

3.6.3.1 CREATE COLLATION

Use the ICU_SHORT_FORM parameter with the CREATE COLLATION command to create
an ICU collation:

CREATE COLLATION collation_name (

[ LOCALE = locale, ]
[ LC_COLLATE = lc_collate, ]
[ LC_CTYPE = lc_ctype, ]
[ ICU_SHORT_FORM = icu_short_form ]
);

To be able to create a collation, you must have CREATE privilege on the destination
schema where the collation is to reside.

For information about the general usage of the CREATE COLLATION command, please
refer to the PostgreSQL core documentation available at:

https://www.postgresql.org/docs/10/static/sql-createcollation.html

UTF-8 character encoding of the database is required. Any LOCALE, or LC_COLLATE and
LC_CTYPE settings that are accepted with UTF-8 encoding can be used.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 154

 EDB Postgres Advanced Server Guide

Parameters

collation_name

The name of the collation to be added. The collation_name may be schema-

qualified.

locale

The locale to be used. Short cut for setting LC_COLLATE and LC_TYPE. If
LOCALE is specified, then LC_COLLATE and LC_TYPE must be omitted.

lc_collate

The collation to be used. If LC_CTYPE is specified, then LC_COLLATE must also

be specified and LOCALE must be omitted.

lc_ctype

The character classification to be used. If LC_COLLATE is specified, then

LC_CTYPE must also be specified and LOCALE must be omitted.

icu_short_form

The text string specifying the collation attributes and their settings. This typically
consists of an ICU short form name, possibly appended with additional collation
attribute/value pairs. A list of ICU short form names is available from column
icu_short_form in system catalog pg_catalog.pg_icu_collate_names.

Example

The following creates a collation using the LROOT ICU short form.

edb=# CREATE COLLATION icu_collate_a (LOCALE = 'en_US.UTF8', ICU_SHORT_FORM =

'LROOT');
CREATE COLLATION

The definition of the new collation can be seen with the following psql command.

edb=# \dO
List of collations
Schema | Name | Collate | Ctype | ICU
--------------+---------------+------------+------------+-------
enterprisedb | icu_collate_a | en_US.UTF8 | en_US.UTF8 | LROOT
(1 row)

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 155

 EDB Postgres Advanced Server Guide

3.6.3.2 CREATE DATABASE

The following is the syntax for creating a database with an ICU collation:

CREATE DATABASE database_name

[ [ WITH ] [ OWNER [=] user_name ]
[ TEMPLATE [=] template ]
[ ENCODING [=] encoding ]
[ LC_COLLATE [=] lc_collate ]
[ LC_CTYPE [=] lc_ctype ]
[ TABLESPACE [=] tablespace_name ]
[ CONNECTION LIMIT [=] connlimit ]
[ ICU_SHORT_FORM [=] icu_short_form ]];

For complete information about the general usage, syntax, and parameters of the CREATE
DATABASE command, please refer to the PostgreSQL core documentation available at:

https://www.postgresql.org/docs/10/static/sql-createdatabase.html

When using the CREATE DATABASE command to create a database using an ICU
collation, the TEMPLATE template0 clause must be specified and the database
encoding must be UTF-8.

The following is an example of creating a database using the LROOT ICU short form
collation, but sorts an uppercase form of a letter before its lowercase counterpart (CU) and
treats variable characters as non-ignorable (AN).

CREATE DATABASE collation_db

TEMPLATE template0
ENCODING 'UTF8'
ICU_SHORT_FORM = 'AN_CU_EX_NX_LROOT';

The following psql command shows the newly created database.

edb=# \l collation_db
List of databases
Name | Owner | Encoding | Collate | Ctype | ICU |
Access privileges
--------------+--------------+----------+-------------+-------------+-------------------+-----
--------------
collation_db | enterprisedb | UTF8 | en_US.UTF-8 | en_US.UTF-8 | AN_CU_EX_NX_LROOT |
(1 row)

The following table is created and populated with rows in the database.

CREATE TABLE collate_tbl (

id INTEGER,
c2 VARCHAR(2)
);

INSERT INTO collate_tbl VALUES (1, 'A');

INSERT INTO collate_tbl VALUES (2, 'B');
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 156
 EDB Postgres Advanced Server Guide

INSERT INTO collate_tbl VALUES (3, 'C');

INSERT INTO collate_tbl VALUES (4, 'a');
INSERT INTO collate_tbl VALUES (5, 'b');
INSERT INTO collate_tbl VALUES (6, 'c');
INSERT INTO collate_tbl VALUES (7, '1');
INSERT INTO collate_tbl VALUES (8, '2');
INSERT INTO collate_tbl VALUES (9, '.B');
INSERT INTO collate_tbl VALUES (10, '-B');
INSERT INTO collate_tbl VALUES (11, ' B');

The following query shows that the uppercase form of a letter sorts before the lowercase
form of the same base letter, and in addition, variable characters are taken into account
when sorted as they appear at the beginning of the sort list. (The default behavior for
en_US.UTF-8 is to sort the lowercase form of a letter before the uppercase form of the
same base letter, and to ignore variable characters.)

collation_db=# SELECT id, c2 FROM collate_tbl ORDER BY c2;

id | c2
----+----
11 | B
10 | -B
9 | .B
7 | 1
8 | 2
1 | A
4 | a
2 | B
5 | b
3 | C
6 | c
(11 rows)

3.6.3.3 initdb
A database cluster can be created with a default ICU collation for all databases in the
cluster by using the --icu-short-form option with the initdb command.

For complete information about the general usage, syntax, and parameters of the initdb
command, please refer to the PostgreSQL core documentation available at:

https://www.postgresql.org/docs/10/static/app-initdb.html

The following illustrates this process.

$ su enterprisedb
Password:
$ /opt/edb/as10/bin/initdb -U enterprisedb -D /tmp/collation_data --encoding
UTF8 --icu-short-form 'AN_CU_EX_NX_LROOT'
The files belonging to this database system will be owned by user
"enterprisedb".
This user must also own the server process.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 157

 EDB Postgres Advanced Server Guide

The database cluster will be initialized with locale "en_US.UTF-8".

The default text search configuration will be set to "english".

Data page checksums are disabled.

creating directory /tmp/collation_data ... ok

creating subdirectories ... ok
.
.
.
Success. You can now start the database server using:

/opt/edb/as10/bin/edb-postgres -D /tmp/collation_data
or
/opt/edb/as10/bin/pg_ctl -D /tmp/collation_data -l logfile start

The following shows the databases created in the cluster which all have an ICU collation
of AN_CU_EX_NX_LROOT.

edb=# \l
List of databases
Name | Owner | Encoding | Collate | Ctype | ICU |
Access privileges
-----------+--------------+----------+-------------+-------------+-------------------+--------
-----------------------
edb | enterprisedb | UTF8 | en_US.UTF-8 | en_US.UTF-8 | AN_CU_EX_NX_LROOT |
postgres | enterprisedb | UTF8 | en_US.UTF-8 | en_US.UTF-8 | AN_CU_EX_NX_LROOT |
template0 | enterprisedb | UTF8 | en_US.UTF-8 | en_US.UTF-8 | AN_CU_EX_NX_LROOT |
=c/enterprisedb +
| | | | | |
enterprisedb=CTc/enterprisedb
template1 | enterprisedb | UTF8 | en_US.UTF-8 | en_US.UTF-8 | AN_CU_EX_NX_LROOT |
=c/enterprisedb +
| | | | | |
enterprisedb=CTc/enterprisedb
(4 rows)

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 158

 EDB Postgres Advanced Server Guide

3.6.4 Using a Collation

A newly defined ICU collation can be used anywhere the COLLATION

"collation_name" clause can be used in a SQL command such as in the column
specifications of the CREATE TABLE command or appended to an expression in the
ORDER BY clause of a SELECT command.

The following are some examples of the creation and usage of ICU collations based on
the English language in the United States (en_US.UTF8).

In these examples, ICU collations are created with the following characteristics.

Collation icu_collate_lowercase forces the lowercase form of a letter to sort before

its uppercase counterpart (CL).

Collation icu_collate_uppercase forces the uppercase form of a letter to sort before

its lowercase counterpart (CU).

Collation icu_collate_ignore_punct causes variable characters (white space and

punctuation marks) to be ignored during sorting (AS).

Collation icu_collate_ignore_white_sp causes white space and other non-visible

variable characters to be ignored during sorting, but visible variable characters
(punctuation marks) are not ignored (AS, T0020).

CREATE COLLATION icu_collate_lowercase (

LOCALE = 'en_US.UTF8',
ICU_SHORT_FORM = 'AN_CL_EX_NX_LROOT'
);

CREATE COLLATION icu_collate_uppercase (

LOCALE = 'en_US.UTF8',
ICU_SHORT_FORM = 'AN_CU_EX_NX_LROOT'
);

CREATE COLLATION icu_collate_ignore_punct (

LOCALE = 'en_US.UTF8',
ICU_SHORT_FORM = 'AS_CX_EX_NX_LROOT_L3'
);

CREATE COLLATION icu_collate_ignore_white_sp (

LOCALE = 'en_US.UTF8',
ICU_SHORT_FORM = 'AS_CX_EX_NX_LROOT_L3_T0020'
);

Note: When creating collations, ICU may generate notice and warning messages when
attributes are given to modify the LROOT collation.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 159

 EDB Postgres Advanced Server Guide

The following psql command lists the collations.

edb=# \dO
List of collations
Schema | Name | Collate | Ctype | ICU
--------------+-----------------------------+------------+------------+-----------------------
-----
enterprisedb | icu_collate_ignore_punct | en_US.UTF8 | en_US.UTF8 | AS_CX_EX_NX_LROOT_L3
enterprisedb | icu_collate_ignore_white_sp | en_US.UTF8 | en_US.UTF8 |
AS_CX_EX_NX_LROOT_L3_T0020
enterprisedb | icu_collate_lowercase | en_US.UTF8 | en_US.UTF8 | AN_CL_EX_NX_LROOT
enterprisedb | icu_collate_uppercase | en_US.UTF8 | en_US.UTF8 | AN_CU_EX_NX_LROOT
(4 rows)

The following table is created and populated.

CREATE TABLE collate_tbl (

id INTEGER,
c2 VARCHAR(2)
);

INSERT INTO collate_tbl VALUES (1, 'A');

INSERT INTO collate_tbl VALUES (2, 'B');
INSERT INTO collate_tbl VALUES (3, 'C');
INSERT INTO collate_tbl VALUES (4, 'a');
INSERT INTO collate_tbl VALUES (5, 'b');
INSERT INTO collate_tbl VALUES (6, 'c');
INSERT INTO collate_tbl VALUES (7, '1');
INSERT INTO collate_tbl VALUES (8, '2');
INSERT INTO collate_tbl VALUES (9, '.B');
INSERT INTO collate_tbl VALUES (10, '-B');
INSERT INTO collate_tbl VALUES (11, ' B');

The following query sorts on column c2 using the default collation. Note that variable
characters (white space and punctuation marks) with id column values of 9, 10, and 11
are ignored and sort with the letter B.

edb=# SELECT * FROM collate_tbl ORDER BY c2;

id | c2
----+----
7 | 1
8 | 2
4 | a
1 | A
5 | b
2 | B
11 | B
10 | -B
9 | .B
6 | c
3 | C
(11 rows)

The following query sorts on column c2 using collation icu_collate_lowercase,

which forces the lowercase form of a letter to sort before the uppercase form of the same
base letter. Also note that the AN attribute forces variable characters to be included in the

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 160

 EDB Postgres Advanced Server Guide

sort order at the same level when comparing base characters so rows with id values of 9,
10, and 11 appear at the beginning of the sort list before all letters and numbers.

edb=# SELECT * FROM collate_tbl ORDER BY c2 COLLATE "icu_collate_lowercase";

id | c2
----+----
11 | B
10 | -B
9 | .B
7 | 1
8 | 2
4 | a
1 | A
5 | b
2 | B
6 | c
3 | C
(11 rows)

The following query sorts on column c2 using collation icu_collate_uppercase,

which forces the uppercase form of a letter to sort before the lowercase form of the same
base letter.

edb=# SELECT * FROM collate_tbl ORDER BY c2 COLLATE "icu_collate_uppercase";

id | c2
----+----
11 | B
10 | -B
9 | .B
7 | 1
8 | 2
1 | A
4 | a
2 | B
5 | b
3 | C
6 | c
(11 rows)

The following query sorts on column c2 using collation

icu_collate_ignore_punct, which causes variable characters to be ignored so rows
with id values of 9, 10, and 11 sort with the letter B as that is the character immediately
following the ignored variable character.

edb=# SELECT * FROM collate_tbl ORDER BY c2 COLLATE

"icu_collate_ignore_punct";
id | c2
----+----
7 | 1
8 | 2
4 | a
1 | A
5 | b
11 | B
2 | B
9 | .B
10 | -B

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 161

 EDB Postgres Advanced Server Guide

6 | c
3 | C
(11 rows)

The following query sorts on column c2 using collation

icu_collate_ignore_white_sp. The AS and T0020 attributes of the collation cause
variable characters with code points less than or equal to hexadecimal 0020 to be ignored
while variable characters with code points greater than hexadecimal 0020 are included in
the sort.

The row with id value of 11, which starts with a space character (hexadecimal 0020)
sorts with the letter B. The rows with id values of 9 and 10, which start with visible
punctuation marks greater than hexadecimal 0020, appear at the beginning of the sort list
as these particular variable characters are included in the sort order at the same level
when comparing base characters.

edb=# SELECT * FROM collate_tbl ORDER BY c2 COLLATE

"icu_collate_ignore_white_sp";
id | c2
----+----
10 | -B
9 | .B
7 | 1
8 | 2
4 | a
1 | A
5 | b
11 | B
2 | B
6 | c
3 | C
(11 rows)

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 162

 EDB Postgres Advanced Server Guide

3.7 Customizable WAL Segment File Size

When a database cluster is created with the initdb utility program, the default size of
each WAL segment file is 16 MB.

The Advanced Server initdb utility provides an additional --wal-segsize option to

specify the size of WAL segment files when creating a new data base cluster.

initdb --wal-segsize=size [ options ]... -D directory

size is the WAL segment file size in megabytes, which must be a power of 2 (for
example, 1, 2, 4, 8, 16, 32, etc.). The minimum permitted value of size is 1 and the
maximum permitted value is 1024. The database cluster is to be created in directory.

For more information on the initdb utility and its other options, see the PostgreSQL
core documentation available at:

https://www.postgresql.org/docs/10/static/app-initdb.html

The following example shows the creation of a database cluster where the WAL segment
file size is specified as 1024 MB (equivalent to 1 GB).

bash-4.1$ initdb -D /opt/data_1024 -U enterprisedb --wal-segsize 1024

The files belonging to this database system will be owned by user
"enterprisedb".
This user must also own the server process.

The database cluster will be initialized with locale "en_US.UTF-8".

The default database encoding has accordingly been set to "UTF8".
The default text search configuration will be set to "english".

Data page checksums are disabled.

fixing permissions on existing directory /opt/data_1024 ... ok

creating subdirectories ... ok
selecting default max_connections ... 100
selecting default shared_buffers ... 128MB
selecting dynamic shared memory implementation ... posix
creating configuration files ... ok
running bootstrap script ... ok
performing post-bootstrap initialization ... ok
creating edb sys ... ok
loading edb contrib modules ...
edb_redwood_bytea.sql ok
edb_redwood_date.sql ok
dbms_alert_public.sql ok
.
.
.
Success. You can now start the database server using:

pg_ctl -D /opt/data_1024 -l logfile start

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 163
 EDB Postgres Advanced Server Guide

The following shows the WAL file as 1.0 GB:

bash-4.1$ pwd
/opt/data_1024/pg_wal
bash-4.1$ ls -lh
total 1.1G
-rw------- 1 enterprisedb enterprisedb 1.0G Jul 24 14:42 000000010000000000000001
drwx------ 2 enterprisedb enterprisedb 4.0K Jul 24 14:42 archive_status

After the database server is started, the display of the wal_segment_size parameter
also confirms the file size:

edb=# SHOW wal_segment_size;

wal_segment_size
------------------
1GB
(1 row)

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 164

 EDB Postgres Advanced Server Guide

4 Security
The chapter describes various features providing for added security.

4.1 Protecting Against SQL Injection Attacks

Advanced Server provides protection against SQL injection attacks. A SQL injection
attack is an attempt to compromise a database by running SQL statements whose results
provide clues to the attacker as to the content, structure, or security of that database.

Preventing a SQL injection attack is normally the responsibility of the application

developer. The database administrator typically has little or no control over the potential
threat. The difficulty for database administrators is that the application must have access
to the data to function properly.

SQL/Protect is a module that allows a database administrator to protect a database from

SQL injection attacks. SQL/Protect provides a layer of security in addition to the normal
database security policies by examining incoming queries for common SQL injection
profiles.

SQL/Protect gives the control back to the database administrator by alerting the
administrator to potentially dangerous queries and by blocking these queries.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 165

 EDB Postgres Advanced Server Guide

4.1.1 SQL/Protect Overview

This section contains an introduction to the different types of SQL injection attacks and
describes how SQL/Protect guards against them.

4.1.1.1 Types of SQL Injection Attacks

There are a number of different techniques used to perpetrate SQL injection attacks. Each
technique is characterized by a certain signature. SQL/Protect examines queries for the
following signatures:

Unauthorized Relations

While Advanced Server allows administrators to restrict access to relations (tables,

views, etc.), many administrators do not perform this tedious task. SQL/Protect provides
a learn mode that tracks the relations a user accesses.

This allows administrators to examine the workload of an application, and for

SQL/Protect to learn which relations an application should be allowed to access for a
given user or group of users in a role.

When SQL/Protect is switched to either passive or active mode, the incoming queries are
checked against the list of learned relations.

Utility Commands

A common technique used in SQL injection attacks is to run utility commands, which are
typically SQL Data Definition Language (DDL) statements. An example is creating a
user-defined function that has the ability to access other system resources.

SQL/Protect can block the running of all utility commands, which are not normally
needed during standard application processing.

SQL Tautology

The most frequent technique used in SQL injection attacks is issuing a tautological
WHERE clause condition (that is, using a condition that is always true).

The following is an example:

WHERE password = 'x' OR 'x'='x'

Attackers will usually start identifying security weaknesses using this technique.
SQL/Protect can block queries that use a tautological conditional clause.
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 166
 EDB Postgres Advanced Server Guide

Unbounded DML Statements

A dangerous action taken during SQL injection attacks is the running of unbounded DML
statements. These are UPDATE and DELETE statements with no WHERE clause. For
example, an attacker may update all users passwords to a known value or initiate a
denial of service attack by deleting all of the data in a key table.

4.1.1.2 Monitoring SQL Injection Attacks

This section describes how SQL/Protect monitors and reports on SQL injection attacks.

4.1.1.2.1 Protected Roles

Monitoring for SQL injection attacks involves analyzing SQL statements originating in
database sessions where the current user of the session is a protected role. A protected
role is an Advanced Server user or group that the database administrator has chosen to
monitor using SQL/Protect. (In Advanced Server, users and groups are collectively
referred to as roles.)

Each protected role can be customized for the types of SQL injection attacks for which it
is to be monitored, thus providing different levels of protection by role and significantly
reducing the user maintenance load for DBAs.

Note: A role with the superuser privilege cannot be made a protected role. If a protected
non-superuser role is subsequently altered to become a superuser, certain behaviors are
exhibited whenever an attempt is made by that superuser to issue any command:

A warning message is issued by SQL/Protect on every command issued by the

protected superuser.
The statistic in column superusers of edb_sql_protect_stats is
incremented with every command issued by the protected superuser. See Section
4.1.1.2.2 for information on the edb_sql_protect_stats view.
When SQL/Protect is in active mode, all commands issued by the protected
superuser are prevented from running.

A protected role that has the superuser privilege should either be altered so that it is no
longer a superuser, or it should be reverted back to an unprotected role.

4.1.1.2.2 Attack Attempt Statistics

Each usage of a command by a protected role that is considered an attack by SQL/Protect

is recorded. Statistics are collected by type of SQL injection attack as discussed in
Section 4.1.1.1.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 167

 EDB Postgres Advanced Server Guide

These statistics are accessible from view edb_sql_protect_stats that can be easily
monitored to identify the start of a potential attack.

The columns in edb_sql_protect_stats monitor the following:

username. Name of the protected role.

superusers. Number of SQL statements issued when the protected role is a
superuser. In effect, any SQL statement issued by a protected superuser increases
this statistic. See Section 4.1.1.2.1 for information on protected superusers.
relations. Number of SQL statements issued referencing relations that were not
learned by a protected role. (That is, relations that are not in a roles protected
relations list.)
commands. Number of DDL statements issued by a protected role.
tautology. Number of SQL statements issued by a protected role that contained a
tautological condition.
dml. Number of UPDATE and DELETE statements issued by a protected role that
did not contain a WHERE clause.

This gives database administrators the opportunity to react proactively in preventing theft
of valuable data or other malicious actions.

If a role is protected in more than one database, the roles statistics for attacks in each
database are maintained separately and are viewable only when connected to the
respective database.

Note: SQL/Protect statistics are maintained in memory while the database server is
running. When the database server is shut down, the statistics are saved to a binary file
named edb_sqlprotect.stat in the data/global subdirectory of the Advanced
Server home directory.

4.1.1.2.3 Attack Attempt Queries

Each usage of a command by a protected role that is considered an attack by SQL/Protect

is recorded in view edb_sql_protect_queries.

View edb_sql_protect_queries contains the following columns:

username. Database user name of the attacker used to log into the database
server.
ip_address. IP address of the machine from which the attack was initiated.
port. Port number from which the attack originated.
machine_name. Name of the machine, if known, from which the attack
originated.
date_time. Date and time at which the query was received by the database server.
The time is stored to the precision of a minute.
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 168
 EDB Postgres Advanced Server Guide

query. The query string sent by the attacker.

The maximum number of offending queries that are saved in

edb_sql_protect_queries is controlled by configuration parameter
edb_sql_protect.max_queries_to_save.

If a role is protected in more than one database, the roles queries for attacks in each
database are maintained separately and are viewable only when connected to the
respective database.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 169

 EDB Postgres Advanced Server Guide

4.1.2 Configuring SQL/Protect

The library file (sqlprotect.so on Linux, sqlprotect.dll on Windows) necessary

to run SQL/Protect should already be installed in the lib subdirectory of your Advanced
Server home directory.

You will also need the SQL script file sqlprotect.sql located in the
share/contrib subdirectory of your Advanced Server home directory.

You must configure the database server to use SQL/Protect, and you must configure each
database that you want SQL/Protect to monitor:

The database server configuration file, postgresql.conf, must be modified by

adding and enabling configuration parameters used by SQL/Protect.
Database objects used by SQL/Protect must be installed in each database that you
want SQL/Protect to monitor.

Step 1: Edit the following configuration parameters in the postgresql.conf file

located in the data subdirectory of your Advanced Server home directory.

shared_preload_libraries. Add $libdir/sqlprotect to the list of libraries.

edb_sql_protect.enabled. Controls whether or not SQL/Protect is actively
monitoring protected roles by analyzing SQL statements issued by those roles and
reacting according to the setting of edb_sql_protect.level. When you are
ready to begin monitoring with SQL/Protect set this parameter to on. If this
parameter is omitted, the default is off.
edb_sql_protect.level. Sets the action taken by SQL/Protect when a SQL
statement is issued by a protected role. If this parameter is omitted, the default
behavior is passive. Initially, set this parameter to learn. See Section 4.1.2.1.2
for further explanation of this parameter.
edb_sql_protect.max_protected_roles. Sets the maximum number of roles that
can be protected. If this parameter is omitted, the default setting is 64. See
Section 3.1.3.12.8 for information on the maximum range of this parameter.
edb_sql_protect.max_protected_relations. Sets the maximum number of
relations that can be protected per role. If this parameter is omitted, the default
setting is 1024.
Please note the total number of protected relations for the server will be the
number of protected relations times the number of protected roles. Every
protected relation consumes space in shared memory. The space for the maximum
possible protected relations is reserved during database server startup.
See Section 3.1.3.12.7 for information about the maximum range of this
parameter.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 170

 EDB Postgres Advanced Server Guide

edb_sql_protect.max_queries_to_save. Sets the maximum number of offending

queries to save in the edb_sql_protect_queries view. If this parameter is
omitted, the default setting is 5000. If the number of offending queries reaches
the limit, additional queries are not saved in the view, but are accessible in the
database server log file. Note: The minimum valid value for this parameter is
100. If a value less than 100 is specified, the database server starts using the
default setting of 5000. A warning message is recorded in the database server log
file. See Section 3.1.3.12.9 for information on the maximum range of this
parameter.

The following example shows the settings of these parameters in the postgresql.conf
file:
shared_preload_libraries = '$libdir/dbms_pipe,$libdir/edb_gen,$libdir/sqlprotect'
# (change requires restart)
.
.
.
edb_sql_protect.enabled = off
edb_sql_protect.level = learn
edb_sql_protect.max_protected_roles = 64
edb_sql_protect.max_protected_relations = 1024
edb_sql_protect.max_queries_to_save = 5000

Step 2: Restart the database server after you have modified the postgresql.conf file.

On Linux: Invoke the Advanced Server service script with the restart option:

On a Redhat or CentOS 6.x installation, use the command:

/etc/init.d/edb-as-10 restart

On a Redhat or CentOS 7.x installation, use the command:

systemctl restart edb-as-10

On Windows: Use the Windows Services applet to restart the service named edb-as-
10.

Step 3: For each database that you want to protect from SQL injection attacks, connect to
the database as a superuser (either enterprisedb or postgres, depending upon your
installation options) and run the script sqlprotect.sql located in the
share/contrib subdirectory of your Advanced Server home directory. The script
creates the SQL/Protect database objects in a schema named sqlprotect.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 171

 EDB Postgres Advanced Server Guide

The following example shows this process to set up protection for a database named edb:

$ /opt/edb/as10/bin/psql -d edb -U enterprisedb

Password for user enterprisedb:
psql.bin (10.0.1)
Type "help" for help.

edb=# \i /opt/edb/as10/share/contrib/sqlprotect.sql
CREATE SCHEMA
GRANT
SET
CREATE TABLE
GRANT
CREATE TABLE
GRANT
CREATE FUNCTION
CREATE FUNCTION
CREATE FUNCTION
CREATE FUNCTION
CREATE FUNCTION
CREATE FUNCTION
CREATE FUNCTION
DO
CREATE FUNCTION
CREATE FUNCTION
DO
CREATE VIEW
GRANT
DO
CREATE VIEW
GRANT
CREATE VIEW
GRANT
CREATE FUNCTION
CREATE FUNCTION
SET

4.1.2.1 Selecting Roles to Protect

After the SQL/Protect database objects have been created in a database, you select the
roles for which SQL queries are to be monitored for protection, and the level of
protection.

4.1.2.1.1 Setting the Protected Roles List

For each database that you want to protect, you must determine the roles you want to
monitor and then add those roles to the protected roles list of that database.

Step 1: Connect as a superuser to a database that you wish to protect using either psql or
Postgres Enterprise Manager Client.

$ /opt/edb/as10/bin/psql -d edb -U enterprisedb

Password for user enterprisedb:
psql.bin (10.0.1)
Type "help" for help.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 172

 EDB Postgres Advanced Server Guide

edb=#

Step 2: Since the SQL/Protect tables, functions, and views are built under the
sqlprotect schema, use the SET search_path command to include the
sqlprotect schema in your search path. This eliminates the need to schema-qualify
any operation or query involving SQL/Protect database objects.

edb=# SET search_path TO sqlprotect;

SET

Step 3: Each role that you wish to protect must be added to the protected roles list. This
list is maintained in the table edb_sql_protect.

To add a role, use the function protect_role('rolename').

The following example protects a role named appuser.

edb=# SELECT protect_role('appuser');

protect_role
--------------

(1 row)

You can list the roles that have been added to the protected roles list by issuing the
following query:
edb=# SELECT * FROM edb_sql_protect;
dbid | roleid | protect_relations | allow_utility_cmds | allow_tautology | allow_empty_dml
-------+--------+-------------------+--------------------+-----------------+-----------------
13917 | 16671 | t | f | f | f
(1 row)

A view is also provided that gives the same information using the object names instead of
the Object Identification numbers (OIDs).

edb=# \x
Expanded display is on.
edb=# SELECT * FROM list_protected_users;
-[ RECORD 1 ]------+--------
dbname | edb
username | appuser
protect_relations | t
allow_utility_cmds | f
allow_tautology | f
allow_empty_dml | f

4.1.2.1.2 Setting the Protection Level

Configuration parameter edb_sql_protect.level sets the protection level, which

defines the behavior of SQL/Protect when a protected role issues a SQL statement. The
defined behavior applies to all roles in the protected roles lists of all databases
configured with SQL/Protect in the database server.
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 173
 EDB Postgres Advanced Server Guide

In the postgresql.conf file the edb_sql_protect.level configuration parameter

can be set to one of the following values to use either learn mode, passive mode, or active
mode:

learn. Tracks the activities of protected roles and records the relations used by the
roles. This is used when initially configuring SQL/Protect so the expected
behaviors of the protected applications are learned.
passive. Issues warnings if protected roles are breaking the defined rules, but does
not stop any SQL statements from executing. This is the next step after
SQL/Protect has learned the expected behavior of the protected roles. This
essentially behaves in intrusion detection mode and can be run in production
when properly monitored.
active. Stops all invalid statements for a protected role. This behaves as a SQL
firewall preventing dangerous queries from running. This is particularly effective
against early penetration testing when the attacker is trying to determine the
vulnerability point and the type of database behind the application. Not only does
SQL/Protect close those vulnerability points, but it tracks the blocked queries
allowing administrators to be alerted before the attacker finds an alternate method
of penetrating the system.

If the edb_sql_protect.level parameter is not set or is omitted from the

configuration file, the default behavior of SQL/Protect is passive.

If you are using SQL/Protect for the first time, set edb_sql_protect.level to
learn.

4.1.2.2 Monitoring Protected Roles

Once you have configured SQL/Protect in a database, added roles to the protected roles
list, and set the desired protection level, you can then activate SQL/Protect in one of learn
mode, passive mode, or active mode. You can then start running your applications.

With a new SQL/Protect installation, the first step is to determine the relations that
protected roles should be permitted to access during normal operation. Learn mode
allows a role to run applications during which time SQL/Protect is recording the relations
that are accessed. These are added to the roles protected relations list stored in table
edb_sql_protect_rel.

Monitoring for protection against attack begins when SQL/Protect is run in passive or
active mode. In passive and active modes, the role is permitted to access the relations in
its protected relations list as these were determined to be the relations the role should be
able to access during typical usage.

However, if a role attempts to access a relation that is not in its protected relations list, a
WARNING or ERROR severity level message is returned by SQL/Protect. The roles

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 174

 EDB Postgres Advanced Server Guide

attempted action on the relation may or may not be carried out depending upon whether
the mode is passive or active.

4.1.2.2.1 Learn Mode

Step 1: To activate SQL/Protect in learn mode, set the following parameters in the
postgresql.conf file as shown below:

edb_sql_protect.enabled = on
edb_sql_protect.level = learn

Step 2: Reload the postgresql.conf file.

Choose Expert Configuration, then Reload Configuration from the Advanced Server
application menu.

Note: For an alternative method of reloading the configuration file, use the
pg_reload_conf function. Be sure you are connected to a database as a superuser and
execute function pg_reload_conf as shown by the following example:

edb=# SELECT pg_reload_conf();

pg_reload_conf
----------------
t
(1 row)

Step 3: Allow the protected roles to run their applications.

As an example the following queries are issued in the psql application by protected role
appuser:

edb=> SELECT * FROM dept;

NOTICE: SQLPROTECT: Learned relation: 16384
deptno | dname | loc
--------+------------+----------
10 | ACCOUNTING | NEW YORK
20 | RESEARCH | DALLAS
30 | SALES | CHICAGO
40 | OPERATIONS | BOSTON
(4 rows)

edb=> SELECT empno, ename, job FROM emp WHERE deptno = 10;
NOTICE: SQLPROTECT: Learned relation: 16391
empno | ename | job
-------+--------+-----------
7782 | CLARK | MANAGER
7839 | KING | PRESIDENT
7934 | MILLER | CLERK
(3 rows)

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 175

 EDB Postgres Advanced Server Guide

SQL/Protect generates a NOTICE severity level message indicating the relation has been
added to the roles protected relations list.

In SQL/Protect learn mode, SQL statements that are cause for suspicion are not prevented
from executing, but a message is issued to alert the user to potentially dangerous
statements as shown by the following example:

edb=> CREATE TABLE appuser_tab (f1 INTEGER);

NOTICE: SQLPROTECT: This command type is illegal for this user
CREATE TABLE
edb=> DELETE FROM appuser_tab;
NOTICE: SQLPROTECT: Learned relation: 16672
NOTICE: SQLPROTECT: Illegal Query: empty DML
DELETE 0

Step 4: As a protected role runs applications, the SQL/Protect tables can be queried to
observe the addition of relations to the roles protected relations list.

Connect as a superuser to the database you are monitoring and set the search path to
include the sqlprotect schema.

edb=# SET search_path TO sqlprotect;

SET

Query the edb_sql_protect_rel table to see the relations added to the protected
relations list:

edb=# SELECT * FROM edb_sql_protect_rel;

dbid | roleid | relid
-------+--------+-------
13917 | 16671 | 16384
13917 | 16671 | 16391
13917 | 16671 | 16672
(3 rows)

The view list_protected_rels is provided that gives more comprehensive

information along with the object names instead of the OIDs.

edb=# SELECT * FROM list_protected_rels;

Database | Protected User | Schema | Name | Type | Owner
----------+----------------+--------+-------------+-------+--------------
edb | appuser | public | dept | Table | enterprisedb
edb | appuser | public | emp | Table | enterprisedb
edb | appuser | public | appuser_tab | Table | appuser
(3 rows)

4.1.2.2.2 Passive Mode

Once you have determined that a roles applications have accessed all relations they will
need, you can now change the protection level so that SQL/Protect can actively monitor
the incoming SQL queries and protect against SQL injection attacks.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 176

 EDB Postgres Advanced Server Guide

Passive mode is the less restrictive of the two protection modes, passive and active.

Step 1: To activate SQL/Protect in passive mode, set the following parameters in the
postgresql.conf file as shown below:

edb_sql_protect.enabled = on
edb_sql_protect.level = passive

Step 2: Reload the configuration file as shown in Step 2 of Section 4.1.2.2.1.

Now SQL/Protect is in passive mode. For relations that have been learned such as the
dept and emp tables of the prior examples, SQL statements are permitted with no special
notification to the client by SQL/Protect as shown by the following queries run by user
appuser:

edb=> SELECT * FROM dept;

deptno | dname | loc
--------+------------+----------
10 | ACCOUNTING | NEW YORK
20 | RESEARCH | DALLAS
30 | SALES | CHICAGO
40 | OPERATIONS | BOSTON
(4 rows)

edb=> SELECT empno, ename, job FROM emp WHERE deptno = 10;
empno | ename | job
-------+--------+-----------
7782 | CLARK | MANAGER
7839 | KING | PRESIDENT
7934 | MILLER | CLERK
(3 rows)

SQL/Protect does not prevent any SQL statement from executing, but issues a message of
WARNING severity level for SQL statements executed against relations that were not
learned, or for SQL statements that contain a prohibited signature as shown in the
following example:

edb=> CREATE TABLE appuser_tab_2 (f1 INTEGER);

WARNING: SQLPROTECT: This command type is illegal for this user
CREATE TABLE
edb=> INSERT INTO appuser_tab_2 VALUES (1);
WARNING: SQLPROTECT: Illegal Query: relations
INSERT 0 1
edb=> INSERT INTO appuser_tab_2 VALUES (2);
WARNING: SQLPROTECT: Illegal Query: relations
INSERT 0 1
edb=> SELECT * FROM appuser_tab_2 WHERE 'x' = 'x';
WARNING: SQLPROTECT: Illegal Query: relations
WARNING: SQLPROTECT: Illegal Query: tautology
f1
----
1
2
(2 rows)

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 177

 EDB Postgres Advanced Server Guide

Step 3: Monitor the statistics for suspicious activity.

By querying the view edb_sql_protect_stats, you can see the number of times
SQL statements were executed that referenced relations that were not in a roles protected
relations list, or contained SQL injection attack signatures. See Section 4.1.1.2.2 for more
information on view edb_sql_protect_stats.

The following is a query on edb_sql_protect_stats:

edb=# SET search_path TO sqlprotect;

SET
edb=# SELECT * FROM edb_sql_protect_stats;
username | superusers | relations | commands | tautology | dml
----------+------------+-----------+----------+-----------+-----
appuser | 0 | 3 | 1 | 1 | 0
(1 row)

Step 4: View information on specific attacks.

By querying the view edb_sql_protect_queries, you can see the SQL statements
that were executed that referenced relations that were not in a roles protected relations
list, or contained SQL injection attack signatures. See Section 4.1.1.2.3 for more
information on view edb_sql_protect_queries.

The following is a query on edb_sql_protect_queries:

edb=# SELECT * FROM edb_sql_protect_queries;

-[ RECORD 1 ]+---------------------------------------------
username | appuser
ip_address |
port |
machine_name |
date_time | 20-JUN-14 13:21:00 -04:00
query | INSERT INTO appuser_tab_2 VALUES (1);
-[ RECORD 2 ]+---------------------------------------------
username | appuser
ip_address |
port |
machine_name |
date_time | 20-JUN-14 13:21:00 -04:00
query | CREATE TABLE appuser_tab_2 (f1 INTEGER);
-[ RECORD 3 ]+---------------------------------------------
username | appuser
ip_address |
port |
machine_name |
date_time | 20-JUN-14 13:22:00 -04:00
query | INSERT INTO appuser_tab_2 VALUES (2);
-[ RECORD 4 ]+---------------------------------------------
username | appuser
ip_address |
port |
machine_name |
date_time | 20-JUN-14 13:22:00 -04:00
query | SELECT * FROM appuser_tab_2 WHERE 'x' = 'x';

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 178

 EDB Postgres Advanced Server Guide

Note: The ip_address and port columns do not return any information if the attack
originated on the same host as the database server using the Unix-domain socket (that is,
pg_hba.conf connection type local).

4.1.2.2.3 Active Mode

In active mode, disallowed SQL statements are prevented from executing. Also, the
message issued by SQL/Protect has a higher severity level of ERROR instead of WARNING.

Step 1: To activate SQL/Protect in active mode, set the following parameters in the
postgresql.conf file as shown below:

edb_sql_protect.enabled = on
edb_sql_protect.level = active

Step 2: Reload the configuration file as shown in Step 2 of Section 4.1.2.2.1.

The following example illustrates SQL statements similar to those given in the examples
of Step 2 in Section 4.1.2.2.2, but executed by user appuser when
edb_sql_protect.level is set to active:

edb=> CREATE TABLE appuser_tab_3 (f1 INTEGER);

ERROR: SQLPROTECT: This command type is illegal for this user
edb=> INSERT INTO appuser_tab_2 VALUES (1);
ERROR: SQLPROTECT: Illegal Query: relations
edb=> SELECT * FROM appuser_tab_2 WHERE 'x' = 'x';
ERROR: SQLPROTECT: Illegal Query: relations

The following shows the resulting statistics:

edb=# SELECT * FROM sqlprotect.edb_sql_protect_stats;

username | superusers | relations | commands | tautology | dml
----------+------------+-----------+----------+-----------+-----
appuser | 0 | 5 | 2 | 1 | 0
(1 row)

The following is a query on edb_sql_protect_queries:

edb=# SELECT * FROM sqlprotect.edb_sql_protect_queries;

-[ RECORD 1 ]+---------------------------------------------
username | appuser
ip_address |
port |
machine_name |
date_time | 20-JUN-14 13:21:00 -04:00
query | CREATE TABLE appuser_tab_2 (f1 INTEGER);
-[ RECORD 2 ]+---------------------------------------------
username | appuser
ip_address |
port |
machine_name |
date_time | 20-JUN-14 13:22:00 -04:00
query | INSERT INTO appuser_tab_2 VALUES (2);
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 179
 EDB Postgres Advanced Server Guide

-[ RECORD 3 ]+---------------------------------------------
username | appuser
ip_address | 192.168.2.6
port | 50098
machine_name |
date_time | 20-JUN-14 13:39:00 -04:00
query | CREATE TABLE appuser_tab_3 (f1 INTEGER);
-[ RECORD 4 ]+---------------------------------------------
username | appuser
ip_address | 192.168.2.6
port | 50098
machine_name |
date_time | 20-JUN-14 13:39:00 -04:00
query | INSERT INTO appuser_tab_2 VALUES (1);
-[ RECORD 5 ]+---------------------------------------------
username | appuser
ip_address | 192.168.2.6
port | 50098
machine_name |
date_time | 20-JUN-14 13:39:00 -04:00
query | SELECT * FROM appuser_tab_2 WHERE 'x' = 'x';

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 180

 EDB Postgres Advanced Server Guide

4.1.3 Common Maintenance Operations

The following describes how to perform other common operations.

You must be connected as a superuser to perform these operations and have included
schema sqlprotect in your search path.

4.1.3.1 Adding a Role to the Protected Roles List

To add a role to the protected roles list run protect_role('rolename').

protect_role('rolename')

This is shown by the following example:

edb=# SELECT protect_role('newuser');

protect_role
--------------

(1 row)

4.1.3.2 Removing a Role From the Protected Roles List

To remove a role from the protected roles list use either of the following functions:

unprotect_role('rolename')
unprotect_role(roleoid)

Note: The variation of the function using the OID is useful if you remove the role using
the DROP ROLE or DROP USER SQL statement before removing the role from the
protected roles list. If a query on a SQL/Protect relation returns a value such as unknown
(OID=16458) for the user name, use the unprotect_role(roleoid) form of the
function to remove the entry for the deleted role from the protected roles list.

Removing a role using these functions also removes the roles protected relations list.

The statistics for a role that has been removed are not deleted until you use the
drop_stats function as described in Section 4.1.3.5.

The offending queries for a role that has been removed are not deleted until you use the
drop_queries function as described in Section 4.1.3.6.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 181

 EDB Postgres Advanced Server Guide

The following is an example of the unprotect_role function:

edb=# SELECT unprotect_role('newuser');

unprotect_role
----------------

(1 row)

Alternatively, the role could be removed by giving its OID of 16693:

edb=# SELECT unprotect_role(16693);

unprotect_role
----------------

(1 row)

4.1.3.3 Setting the Types of Protection for a Role

You can change whether or not a role is protected from a certain type of SQL injection
attack.

Change the Boolean value for the column in edb_sql_protect corresponding to the
type of SQL injection attack for which protection of a role is to be disabled or enabled.

Be sure to qualify the following columns in your WHERE clause of the statement that
updates edb_sql_protect:

dbid. OID of the database for which you are making the change
roleid. OID of the role for which you are changing the Boolean settings

For example, to allow a given role to issue utility commands, update the
allow_utility_cmds column as follows:

UPDATE edb_sql_protect SET allow_utility_cmds = TRUE WHERE dbid = 13917 AND

roleid = 16671;

You can verify the change was made by querying edb_sql_protect or

list_protected_users. In the following query note that column
allow_utility_cmds now contains t.

edb=# SELECT dbid, roleid, allow_utility_cmds FROM edb_sql_protect;

dbid | roleid | allow_utility_cmds
-------+--------+--------------------
13917 | 16671 | t
(1 row)

The updated rules take effect on new sessions started by the role since the change was
made.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 182

 EDB Postgres Advanced Server Guide

4.1.3.4 Removing a Relation From the Protected Relations List

If SQL/Protect has learned that a given relation is accessible for a given role, you can
subsequently remove that relation from the roles protected relations list.

Delete its entry from the edb_sql_protect_rel table using any of the following
functions:

unprotect_rel('rolename', 'relname')
unprotect_rel('rolename', 'schema', 'relname')
unprotect_rel(roleoid, reloid)

If the relation given by relname is not in your current search path, specify the relations
schema using the second function format.

The third function format allows you to specify the OIDs of the role and relation,
respectively, instead of their text names.

The following example illustrates the removal of the public.emp relation from the
protected relations list of the role appuser.

edb=# SELECT unprotect_rel('appuser', 'public', 'emp');

unprotect_rel
---------------

(1 row)

The following query shows there is no longer an entry for the emp relation.

edb=# SELECT * FROM list_protected_rels;

Database | Protected User | Schema | Name | Type | Owner
----------+----------------+--------+-------------+-------+--------------
edb | appuser | public | dept | Table | enterprisedb
edb | appuser | public | appuser_tab | Table | appuser
(2 rows)

SQL/Protect will now issue a warning or completely block access (depending upon the
setting of edb_sql_protect.level) whenever the role attempts to utilize that
relation.

4.1.3.5 Deleting Statistics

You can delete statistics from view edb_sql_protect_stats using either of the two
following functions:

drop_stats('rolename')
drop_stats(roleoid)

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 183

 EDB Postgres Advanced Server Guide

Note: The variation of the function using the OID is useful if you remove the role using
the DROP ROLE or DROP USER SQL statement before deleting the roles statistics using
drop_stats('rolename'). If a query on edb_sql_protect_stats returns a value
such as unknown (OID=16458) for the user name, use the drop_stats(roleoid)
form of the function to remove the deleted roles statistics from
edb_sql_protect_stats.

The following is an example of the drop_stats function:

edb=# SELECT drop_stats('appuser');

drop_stats
------------

(1 row)

edb=# SELECT * FROM edb_sql_protect_stats;

username | superusers | relations | commands | tautology | dml
----------+------------+-----------+----------+-----------+-----
(0 rows)

The following is an example of using the drop_stats(roleoid) form of the function

when a role is dropped before deleting its statistics:

edb=# SELECT * FROM edb_sql_protect_stats;

username | superusers | relations | commands | tautology | dml
---------------------+------------+-----------+----------+-----------+-----
unknown (OID=16693) | 0 | 5 | 3 | 1 | 0
appuser | 0 | 5 | 2 | 1 | 0
(2 rows)

edb=# SELECT drop_stats(16693);

drop_stats
------------

(1 row)

edb=# SELECT * FROM edb_sql_protect_stats;

username | superusers | relations | commands | tautology | dml
----------+------------+-----------+----------+-----------+-----
appuser | 0 | 5 | 2 | 1 | 0
(1 row)

4.1.3.6 Deleting Offending Queries

You can delete offending queries from view edb_sql_protect_queries using either
of the two following functions:

drop_queries('rolename')
drop_queries(roleoid)

Note: The variation of the function using the OID is useful if you remove the role using
the DROP ROLE or DROP USER SQL statement before deleting the roles offending
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 184
 EDB Postgres Advanced Server Guide

queries using drop_queries('rolename'). If a query on

edb_sql_protect_queries returns a value such as unknown (OID=16454) for the
user name, use the drop_queries(roleoid) form of the function to remove the
deleted roles offending queries from edb_sql_protect_queries.

The following is an example of the drop_queries function:

edb=# SELECT drop_queries('appuser');

drop_queries
--------------
5
(1 row)

edb=# SELECT * FROM edb_sql_protect_queries;

username | ip_address | port | machine_name | date_time | query
----------+------------+------+--------------+-----------+-------
(0 rows)

The following is an example of using the drop_queries(roleoid) form of the

function when a role is dropped before deleting its queries:

edb=# SELECT username, query FROM edb_sql_protect_queries;

username | query
---------------------+----------------------------------------------
unknown (OID=16454) | CREATE TABLE appuser_tab_2 (f1 INTEGER);
unknown (OID=16454) | INSERT INTO appuser_tab_2 VALUES (2);
unknown (OID=16454) | CREATE TABLE appuser_tab_3 (f1 INTEGER);
unknown (OID=16454) | INSERT INTO appuser_tab_2 VALUES (1);
unknown (OID=16454) | SELECT * FROM appuser_tab_2 WHERE 'x' = 'x';
(5 rows)

edb=# SELECT drop_queries(16454);

drop_queries
--------------
5
(1 row)

edb=# SELECT * FROM edb_sql_protect_queries;

username | ip_address | port | machine_name | date_time | query
----------+------------+------+--------------+-----------+-------
(0 rows)

4.1.3.7 Disabling and Enabling Monitoring

If you wish to turn off SQL/Protect monitoring once you have enabled it, perform the
following steps:

Step 1: Set the configuration parameter edb_sql_protect.enabled to off in the

postgresql.conf file.

The entry for edb_sql_protect.enabled should look like the following:

edb_sql_protect.enabled = off
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 185
 EDB Postgres Advanced Server Guide

Step 2: Reload the configuration file as shown in Step 2 of Section 4.1.2.2.1.

To re-enable SQL/Protect monitoring perform the following steps:

Step 1: Set the configuration parameter edb_sql_protect.enabled to on in the

postgresql.conf file.

The entry for edb_sql_protect.enabled should look like the following:

edb_sql_protect.enabled = on

Step 2: Reload the configuration file as shown in Step 2 of Section 4.1.2.2.1.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 186

 EDB Postgres Advanced Server Guide

4.1.4 Backing Up and Restoring a SQL/Protect Database

Backing up a database that is configured with SQL/Protect, and then restoring the backup
file to a new database require additional considerations to what is normally associated
with backup and restore procedures. This is primarily due to the use of Object
Identification numbers (OIDs) in the SQL/Protect tables as explained in this section.

Note: This section is applicable if your backup and restore procedures result in the re-
creation of database objects in the new database with new OIDs such as is the case when
using the pg_dump backup program.

If you are backing up your Advanced Server database server by simply using the
operating systems copy utility to create a binary image of the Advanced Server data files
(file system backup method), then this section does not apply.

4.1.4.1 Object Identification Numbers in SQL/Protect Tables

SQL/Protect uses two tables, edb_sql_protect and edb_sql_protect_rel, to store

information on database objects such as databases, roles, and relations. References to
these database objects in these tables are done using the objects OIDs, and not the
objects text names. The OID is a numeric data type used by Advanced Server to
uniquely identify each database object.

When a database object is created, Advanced Server assigns an OID to the object, which
is then used whenever a reference is needed to the object in the database catalogs. If you
create the same database object in two databases, such as a table with the same CREATE
TABLE statement, each table is assigned a different OID in each database.

In a backup and restore operation that results in the re-creation of the backed up database
objects, the restored objects end up with different OIDs in the new database than what
they were assigned in the original database. As a result, the OIDs referencing databases,
roles, and relations stored in the edb_sql_protect and edb_sql_protect_rel
tables are no longer valid when these tables are simply dumped to a backup file and then
restored to a new database.

The following sections describe two functions, export_sqlprotect and

import_sqlprotect, that are used specifically for backing up and restoring
SQL/Protect tables in order to ensure the OIDs in the SQL/Protect tables reference the
correct database objects after the SQL/Protect tables are restored.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 187

 EDB Postgres Advanced Server Guide

4.1.4.2 Backing Up the Database

The following are the steps to back up a database that has been configured with
SQL/Protect.

Step 1: Create a backup file using pg_dump.

The following example shows a plain-text backup file named /tmp/edb.dmp created
from database edb using the pg_dump utility program:

$ cd /opt/edb/as10/bin
$ ./pg_dump -U enterprisedb -Fp -f /tmp/edb.dmp edb
Password:
$

Step 2: Connect to the database as a superuser and export the SQL/Protect data using the
export_sqlprotect('sqlprotect_file') function where sqlprotect_file is
the fully qualified path to a file where the SQL/Protect data is to be saved.

The enterprisedb operating system account (postgres if you installed Advanced

Server in PostgreSQL compatibility mode) must have read and write access to the
directory specified in sqlprotect_file.

edb=# SELECT sqlprotect.export_sqlprotect('/tmp/sqlprotect.dmp');

export_sqlprotect
-------------------

(1 row)

The files /tmp/edb.dmp and /tmp/sqlprotect.dmp comprise your total database

backup.

4.1.4.3 Restoring From the Backup Files

Step 1: Restore the backup file to the new database.

The following example uses the psql utility program to restore the plain-text backup file
/tmp/edb.dmp to a newly created database named newdb:

$ /opt/edb/as10/bin/psql -d newdb -U enterprisedb -f /tmp/edb.dmp

Password for user enterprisedb:
SET
SET
SET
SET
SET
COMMENT
CREATE SCHEMA
.
.
.
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 188
 EDB Postgres Advanced Server Guide

Step 2: Connect to the new database as a superuser and delete all rows from the
edb_sql_protect_rel table.

This step removes any existing rows in the edb_sql_protect_rel table that were
backed up from the original database. These rows do not contain the correct OIDs
relative to the database where the backup file has been restored.

$ /opt/edb/as10/bin/psql -d newdb -U enterprisedb

Password for user enterprisedb:
psql.bin (10.0.1)
Type "help" for help.

newdb=# DELETE FROM sqlprotect.edb_sql_protect_rel;

DELETE 2

Step 3: Delete all rows from the edb_sql_protect table.

This step removes any existing rows in the edb_sql_protect table that were backed
up from the original database. These rows do not contain the correct OIDs relative to the
database where the backup file has been restored.

newdb=# DELETE FROM sqlprotect.edb_sql_protect;

DELETE 1

Step 4: Delete any statistics that may exist for the database.

This step removes any existing statistics that may exist for the database to which you are
restoring the backup. The following query displays any existing statistics:

newdb=# SELECT * FROM sqlprotect.edb_sql_protect_stats;

username | superusers | relations | commands | tautology | dml
----------+------------+-----------+----------+-----------+-----
(0 rows)

For each row that appears in the preceding query, use the drop_stats function
specifying the role name of the entry.

For example, if a row appeared with appuser in the username column, issue the
following command to remove it:

newdb=# SELECT sqlprotect.drop_stats('appuser');

drop_stats
------------

(1 row)

Step 5: Delete any offending queries that may exist for the database.

This step removes any existing queries that may exist for the database to which you are
restoring the backup. The following query displays any existing queries:
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 189
 EDB Postgres Advanced Server Guide

edb=# SELECT * FROM sqlprotect.edb_sql_protect_queries;

username | ip_address | port | machine_name | date_time | query
----------+------------+------+--------------+-----------+-------
(0 rows)

For each row that appears in the preceding query, use the drop_queries function
specifying the role name of the entry.

For example, if a row appeared with appuser in the username column, issue the
following command to remove it:

edb=# SELECT sqlprotect.drop_queries('appuser');

drop_queries
--------------

(1 row)

Step 6: Make sure the role names that were protected by SQL/Protect in the original
database exist in the database server where the new database resides.

If the original and new databases reside in the same database server, then nothing needs
to be done assuming you have not deleted any of these roles from the database server.

Step 7: Run the function import_sqlprotect('sqlprotect_file') where

sqlprotect_file is the fully qualified path to the file you created in Step 2 of Section
4.1.4.2.

newdb=# SELECT sqlprotect.import_sqlprotect('/tmp/sqlprotect.dmp');

import_sqlprotect
-------------------

(1 row)

Tables edb_sql_protect and edb_sql_protect_rel are now populated with

entries containing the OIDs of the database objects as assigned in the new database. The
statistics view edb_sql_protect_stats also now displays the statistics imported
from the original database.

The SQL/Protect tables and statistics are now properly restored for this database. This is
verified by the following queries on the Advanced Server system catalogs:
newdb=# SELECT datname, oid FROM pg_database;
datname | oid
-----------+-------
template1 | 1
template0 | 13909
edb | 13917
newdb | 16679
(4 rows)

newdb=# SELECT rolname, oid FROM pg_roles;

rolname | oid
--------------+-------
enterprisedb | 10
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 190
 EDB Postgres Advanced Server Guide
appuser | 16671
newuser | 16678
(3 rows)

newdb=# SELECT relname, oid FROM pg_class WHERE relname IN ('dept','emp','appuser_tab');

relname | oid
-------------+-------
appuser_tab | 16803
dept | 16809
emp | 16812
(3 rows)

newdb=# SELECT * FROM sqlprotect.edb_sql_protect;

dbid | roleid | protect_relations | allow_utility_cmds | allow_tautology | allow_empty_dml
-------+--------+-------------------+--------------------+-----------------+-----------------
16679 | 16671 | t | t | f | f
(1 row)

newdb=# SELECT * FROM sqlprotect.edb_sql_protect_rel;

dbid | roleid | relid
-------+--------+-------
16679 | 16671 | 16809
16679 | 16671 | 16803
(2 rows)

newdb=# SELECT * FROM sqlprotect.edb_sql_protect_stats;

username | superusers | relations | commands | tautology | dml
----------+------------+-----------+----------+-----------+-----
appuser | 0 | 5 | 2 | 1 | 0
(1 row)

newedb=# \x
Expanded display is on.
nwedb=# SELECT * FROM sqlprotect.edb_sql_protect_queries;
-[ RECORD 1 ]+---------------------------------------------
username | appuser
ip_address |
port |
machine_name |
date_time | 20-JUN-14 13:21:00 -04:00
query | CREATE TABLE appuser_tab_2 (f1 INTEGER);
-[ RECORD 2 ]+---------------------------------------------
username | appuser
ip_address |
port |
machine_name |
date_time | 20-JUN-14 13:22:00 -04:00
query | INSERT INTO appuser_tab_2 VALUES (2);
-[ RECORD 3 ]+---------------------------------------------
username | appuser
ip_address | 192.168.2.6
port | 50098
machine_name |
date_time | 20-JUN-14 13:39:00 -04:00
query | CREATE TABLE appuser_tab_3 (f1 INTEGER);
-[ RECORD 4 ]+---------------------------------------------
username | appuser
ip_address | 192.168.2.6
port | 50098
machine_name |
date_time | 20-JUN-14 13:39:00 -04:00
query | INSERT INTO appuser_tab_2 VALUES (1);
-[ RECORD 5 ]+---------------------------------------------
username | appuser
ip_address | 192.168.2.6
port | 50098
machine_name |
date_time | 20-JUN-14 13:39:00 -04:00
query | SELECT * FROM appuser_tab_2 WHERE 'x' = 'x';

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 191

 EDB Postgres Advanced Server Guide

Note the following about the columns in tables edb_sql_protect and

edb_sql_protect_rel:

dbid. Matches the value in the oid column from pg_database for newdb
roleid. Matches the value in the oid column from pg_roles for appuser

Also note that in table edb_sql_protect_rel, the values in the relid column match
the values in the oid column of pg_class for relations dept and appuser_tab.

Step 8: Verify that the SQL/Protect configuration parameters are set as desired in the
postgresql.conf file for the database server running the new database. Restart the
database server or reload the configuration file as appropriate.

You can now monitor the database using SQL/Protect.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 192

 EDB Postgres Advanced Server Guide

4.2 Virtual Private Database

Virtual Private Database is a type of fine-grained access control using security policies.
Fine-grained access control in Virtual Private Database means that access to data can be
controlled down to specific rows as defined by the security policy.

The rules that encode a security policy are defined in a policy function, which is an SPL
function with certain input parameters and return value. The security policy is the named
association of the policy function to a particular database object, typically a table.

Note: In Advanced Server, the policy function can be written in any language supported
by Advanced Server such as SQL and PL/pgSQL in addition to SPL.

Note: The database objects currently supported by Advanced Server Virtual Private
Database are tables. Policies cannot be applied to views or synonyms.

The advantages of using Virtual Private Database are the following:

Provides a fine-grained level of security. Database object level privileges given by

the GRANT command determine access privileges to the entire instance of a
database object, while Virtual Private Database provides access control for the
individual rows of a database object instance.
A different security policy can be applied depending upon the type of SQL
command (INSERT, UPDATE, DELETE, or SELECT).
The security policy can vary dynamically for each applicable SQL command
affecting the database object depending upon factors such as the session user of
the application accessing the database object.
Invocation of the security policy is transparent to all applications that access the
database object and thus, individual applications do not have to be modified to
apply the security policy.
Once a security policy is enabled, it is not possible for any application (including
new applications) to circumvent the security policy except by the system privilege
noted by the following.
Even superusers cannot circumvent the security policy except by the system
privilege noted by the following.

Note: The only way security policies can be circumvented is if the EXEMPT ACCESS
POLICY system privilege has been granted to a user. The EXEMPT ACCESS POLICY
privilege should be granted with extreme care as a user with this privilege is exempted
from all policies in the database.

The DBMS_RLS package provides procedures to create policies, remove policies, enable
policies, and disable policies.
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 193
 EDB Postgres Advanced Server Guide

4.3 sslutils

sslutils is a Postgres extension that provides SSL certificate generation functions to

Advanced Server for use by the EDB Postgres Enterprise Manager server. sslutils is
installed by the Advanced Server RPM Meta Installer and the Advanced Server graphical
installer.

The sslutils package provides the functions shown in the following sections.

In these sections, each parameter in the functions parameter list is described by

parameter n under the Parameters subsection where n refers to the nth ordinal
position (for example, first, second, third, etc.) within the functions parameter list.

4.3.1 openssl_rsa_generate_key

The openssl_rsa_generate_key function generates an RSA private key. The

function signature is:

openssl_rsa_generate_key(integer) RETURNS text

When invoking the function, pass the number of bits as an integer value; the function
returns the generated key.

4.3.2 openssl_rsa_key_to_csr

The openssl_rsa_key_to_csr function generates a certificate signing request (CSR).

The signature is:

openssl_rsa_key_to_csr(text, text, text, text, text, text,

text) RETURNS text

The function generates and returns the certificate signing request.

Parameters

parameter 1

The name of the RSA key file.

parameter 2

The common name (e.g., agentN) of the agent that will use the signing request.

parameter 3
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 194
 EDB Postgres Advanced Server Guide

The name of the country in which the server resides.

parameter 4

The name of the state in which the server resides.

parameter 5

The location (city) within the state in which the server resides.

parameter 6

The name of the organization unit requesting the certificate.

parameter 7

The email address of the user requesting the certificate.

4.3.3 openssl_csr_to_crt

The openssl_csr_to_crt function generates a self-signed certificate or a certificate

authority certificate. The signature is:

openssl_csr_to_crt(text, text, text) RETURNS text

The function returns the self-signed certificate or certificate authority certificate.

Parameters

parameter 1

The name of the certificate signing the request.

parameter 2

The path to the certificate authority certificate, or NULL if generating a certificate

authority certificate.

parameter 3

The path to the certificate authoritys private key or (if argument 2 is NULL) the
path to a private key.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 195

 EDB Postgres Advanced Server Guide

4.3.4 openssl_rsa_generate_crl

The openssl_rsa_generate_crl function generates a default certificate revocation

list. The signature is:

openssl_rsa_generate_crl(text, text) RETURNS text

The function returns the certificate revocation list.

Parameters

parameter 1

The path to the certificate authority certificate.

parameter 2

The path to the certificate authority private key.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 196

 EDB Postgres Advanced Server Guide

5 EDB Resource Manager

EDB Resource Manager is an Advanced Server feature that provides the capability to
control the usage of operating system resources used by Advanced Server processes.

This capability allows you to protect the system from processes that may uncontrollably
overuse and monopolize certain system resources.

The following are some key points about using EDB Resource Manager.

The basic component of EDB Resource Manager is a resource group. A resource

group is a named, global group, available to all databases in an Advanced Server
instance, on which various resource usage limits can be defined. Advanced Server
processes that are assigned as members of a given resource group are then
controlled by EDB Resource Manager so that the aggregate resource usage of all
processes in the group is kept near the limits defined on the group.
Data definition language commands are used to create, alter, and drop resource
groups. These commands can only be used by a database user with superuser
privileges.
The desired, aggregate consumption level of all processes belonging to a resource
group is defined by resource type parameters. There are different resource type
parameters for the different types of system resources currently supported by
EDB Resource Manager.
Multiple resource groups can be created, each with different settings for its
resource type parameters, thus defining different consumption levels for each
resource group.
EDB Resource Manager throttles processes in a resource group to keep resource
consumption near the limits defined by the resource type parameters. If there are
multiple resource type parameters with defined settings in a resource group, the
actual resource consumption may be significantly lower for certain resource types
than their defined resource type parameter settings. This is because EDB
Resource Manager throttles processes attempting to keep all resources with
defined resource type settings within their defined limits.
The definition of available resource groups and their resource type settings are
stored in a shared global system catalog. Thus, resource groups can be utilized by
all databases in a given Advanced Server instance.
The edb_max_resource_groups configuration parameter sets the maximum
number of resource groups that can be active simultaneously with running
processes. The default setting is 16 resource groups. Changes to this parameter
take effect on database server restart.
Use the SET edb_resource_group TO group_name command to assign the
current process to a specified resource group. Use the RESET
edb_resource_group command or SET edb_resource_group TO
DEFAULT to remove the current process from a resource group.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 197

 EDB Postgres Advanced Server Guide

A default resource group can be assigned to a role using the ALTER ROLE ...
SET command, or to a database by the ALTER DATABASE ... SET command.
The entire database server instance can be assigned a default resource group by
setting the parameter in the postgresql.conf file.
In order to include resource groups in a backup file of the database server
instance, use the pg_dumpall backup utility with default settings (That is, do not
specify any of the --globals-only, --roles-only, or --tablespaces-
only options.)

5.1 Creating and Managing Resource Groups

The data definition language commands described in this section provide for the creation
and management of resource groups.

5.1.1 CREATE RESOURCE GROUP

Use the CREATE RESOURCE GROUP command to create a new resource group.

CREATE RESOURCE GROUP group_name;

Description

The CREATE RESOURCE GROUP command creates a resource group with the specified
name. Resource limits can then be defined on the group with the ALTER RESOURCE
GROUP command. The resource group is accessible from all databases in the Advanced
Server instance.

To use the CREATE RESOURCE GROUP command you must have superuser privileges.

Parameters

group_name

The name of the resource group.

Example

The following example results in the creation of three resource groups named resgrp_a,
resgrp_b, and resgrp_c.

edb=# CREATE RESOURCE GROUP resgrp_a;

CREATE RESOURCE GROUP
edb=# CREATE RESOURCE GROUP resgrp_b;
CREATE RESOURCE GROUP

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 198

 EDB Postgres Advanced Server Guide

edb=# CREATE RESOURCE GROUP resgrp_c;

CREATE RESOURCE GROUP

The following query shows the entries for the resource groups in the
edb_resource_group catalog.

edb=# SELECT * FROM edb_resource_group;

rgrpname | rgrpcpuratelimit | rgrpdirtyratelimit
----------+------------------+--------------------
resgrp_a | 0 | 0
resgrp_b | 0 | 0
resgrp_c | 0 | 0
(3 rows)

5.1.2 ALTER RESOURCE GROUP

Use the ALTER RESOURCE GROUP command to change the attributes of an existing
resource group. The command syntax comes in three forms.

The first form renames the resource group:

ALTER RESOURCE GROUP group_name RENAME TO new_name;

The second form assigns a resource type to the resource group:

ALTER RESOURCE GROUP group_name SET

resource_type { TO | = } { value | DEFAULT };

The third form resets the assignment of a resource type to its default within the group:

ALTER RESOURCE GROUP group_name RESET resource_type;

Description

The ALTER RESOURCE GROUP command changes certain attributes of an existing

resource group.

The first form with the RENAME TO clause assigns a new name to an existing resource
group.

The second form with the SET resource_type TO clause either assigns the specified
literal value to a resource type, or resets the resource type when DEFAULT is specified.
Resetting or setting a resource type to DEFAULT means that the resource group has no
defined limit on that resource type.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 199

 EDB Postgres Advanced Server Guide

The third form with the RESET resource_type clause resets the resource type for the
group as described previously.

To use the ALTER RESOURCE GROUP command you must have superuser privileges.

Parameters

group_name

The name of the resource group to be altered.

new_name

The new name to be assigned to the resource group.

resource_type

The resource type parameter specifying the type of resource to which a usage
value is to be set.

value | DEFAULT

When value is specified, the literal value to be assigned to resource_type.

When DEFAULT is specified, the assignment of resource_type is reset for the
resource group.

Example

The following are examples of the ALTER RESOURCE GROUP command.

edb=# ALTER RESOURCE GROUP resgrp_a RENAME TO newgrp;

ALTER RESOURCE GROUP
edb=# ALTER RESOURCE GROUP resgrp_b SET cpu_rate_limit = .5;
ALTER RESOURCE GROUP
edb=# ALTER RESOURCE GROUP resgrp_b SET dirty_rate_limit = 6144;
ALTER RESOURCE GROUP
edb=# ALTER RESOURCE GROUP resgrp_c RESET cpu_rate_limit;
ALTER RESOURCE GROUP

The following query shows the results of the ALTER RESOURCE GROUP commands to
the entries in the edb_resource_group catalog.

edb=# SELECT * FROM edb_resource_group;

rgrpname | rgrpcpuratelimit | rgrpdirtyratelimit
----------+------------------+--------------------
newgrp | 0 | 0
resgrp_b | 0.5 | 6144
resgrp_c | 0 | 0
(3 rows)

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 200

 EDB Postgres Advanced Server Guide

5.1.3 DROP RESOURCE GROUP

Use the DROP RESOURCE GROUP command to remove a resource group.

DROP RESOURCE GROUP [ IF EXISTS ] group_name;

Description

The DROP RESOURCE GROUP command removes a resource group with the specified
name.

To use the DROP RESOURCE GROUP command you must have superuser privileges.

Parameters

group_name

The name of the resource group to be removed.

IF EXISTS

Do not throw an error if the resource group does not exist. A notice is issued in
this case.

Example

The following example removes resource group newgrp.

edb=# DROP RESOURCE GROUP newgrp;

DROP RESOURCE GROUP

5.1.4 Assigning a Process to a Resource Group

Use the SET edb_resource_group TO group_name command to assign the current

process to a specified resource group as shown by the following.

edb=# SET edb_resource_group TO resgrp_b;

SET
edb=# SHOW edb_resource_group;
edb_resource_group
--------------------
resgrp_b
(1 row)

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 201

 EDB Postgres Advanced Server Guide

The resource type settings of the group immediately take effect on the current process. If
the command is used to change the resource group assigned to the current process, the
resource type settings of the newly assigned group immediately take effect.

Processes can be included by default in a resource group by assigning a default resource

group to roles, databases, or an entire database server instance.

A default resource group can be assigned to a role using the ALTER ROLE ... SET
command. For more information about the ALTER ROLE command, please refer to the
PostgreSQL core documentation available at:

https://www.postgresql.org/docs/10/static/sql-alterrole.html

A default resource group can be assigned to a database by the ALTER DATABASE ...
SET command. For more information about the ALTER DATABASE command, please
refer to the PostgreSQL core documentation available at:

https://www.postgresql.org/docs/10/static/sql-alterdatabase.html

The entire database server instance can be assigned a default resource group by setting
the edb_resource_group configuration parameter in the postgresql.conf file as
shown by the following.

# - EDB Resource Manager -

#edb_max_resource_groups = 16 # 0-65536 (change requires restart)
edb_resource_group = 'resgrp_b'

A change to edb_resource_group in the postgresql.conf file requires a

configuration file reload before it takes effect on the database server instance.

5.1.5 Removing a Process from a Resource Group

Set edb_resource_group to DEFAULT or use RESET edb_resource_group to

remove the current process from a resource group as shown by the following.

edb=# SET edb_resource_group TO DEFAULT;

SET
edb=# SHOW edb_resource_group;
edb_resource_group
--------------------

(1 row)

For removing a default resource group from a role, use the ALTER ROLE ... RESET
form of the ALTER ROLE command.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 202

 EDB Postgres Advanced Server Guide

For removing a default resource group from a database, use the ALTER DATABASE ...
RESET form of the ALTER DATABASE command.

For removing a default resource group from the database server instance, set the
edb_resource_group configuration parameter to an empty string in the
postgresql.conf file and reload the configuration file.

5.1.6 Monitoring Processes in Resource Groups

After resource groups have been created, the number of processes actively using these
resource groups can be obtained from the view edb_all_resource_groups.

The columns in edb_all_resource_groups are the following:

group_name. Name of the resource group.

active_processes. Number of active processes in the resource group.
cpu_rate_limit. The value of the CPU rate limit resource type assigned to the
resource group.
per_process_cpu_rate_limit. The CPU rate limit applicable to an individual,
active process in the resource group.
dirty_rate_limit. The value of the dirty rate limit resource type assigned to the
resource group.
per_process_dirty_rate_limit. The dirty rate limit applicable to an individual,
active process in the resource group.

Note: Columns per_process_cpu_rate_limit and

per_process_dirty_rate_limit do not show the actual resource consumption used
by the processes, but indicate how EDB Resource Manager sets the resource limit for an
individual process based upon the number of active processes in the resource group.

The following shows edb_all_resource_groups when resource group resgrp_a

contains no active processes, resource group resgrp_b contains two active processes,
and resource group resgrp_c contains one active process.

edb=# SELECT * FROM edb_all_resource_groups ORDER BY group_name;

-[ RECORD 1 ]----------------+------------------
group_name | resgrp_a
active_processes | 0
cpu_rate_limit | 0.5
per_process_cpu_rate_limit |
dirty_rate_limit | 12288
per_process_dirty_rate_limit |
-[ RECORD 2 ]----------------+------------------
group_name | resgrp_b
active_processes | 2
cpu_rate_limit | 0.4

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 203

 EDB Postgres Advanced Server Guide

per_process_cpu_rate_limit | 0.195694289022895
dirty_rate_limit | 6144
per_process_dirty_rate_limit | 3785.92924684337
-[ RECORD 3 ]----------------+------------------
group_name | resgrp_c
active_processes | 1
cpu_rate_limit | 0.3
per_process_cpu_rate_limit | 0.292342129631091
dirty_rate_limit | 3072
per_process_dirty_rate_limit | 3072

The CPU rate limit and dirty rate limit settings that are assigned to these resource groups
are as follows.

edb=# SELECT * FROM edb_resource_group;

rgrpname | rgrpcpuratelimit | rgrpdirtyratelimit
----------+------------------+--------------------
resgrp_a | 0.5 | 12288
resgrp_b | 0.4 | 6144
resgrp_c | 0.3 | 3072
(3 rows)

In the edb_all_resource_groups view, note that the

per_process_cpu_rate_limit and per_process_dirty_rate_limit values are
roughly the corresponding CPU rate limit and dirty rate limit divided by the number of
active processes.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 204

 EDB Postgres Advanced Server Guide

5.2 CPU Usage Throttling

CPU usage of a resource group is controlled by setting the cpu_rate_limit resource

type parameter.

Set the cpu_rate_limit parameter to the fraction of CPU time over wall-clock time to
which the combined, simultaneous CPU usage of all processes in the group should not
exceed. Thus, the value assigned to cpu_rate_limit should typically be less than or
equal to 1.

The valid range of the cpu_rate_limit parameter is 0 to 1.67772e+07. A setting of 0

means no CPU rate limit has been set for the resource group.

When multiplied by 100, the cpu_rate_limit can also be interpreted as the CPU usage
percentage for a resource group.

EDB Resource Manager utilizes CPU throttling to keep the aggregate CPU usage of all
processes in the group within the limit specified by the cpu_rate_limit parameter. A
process in the group may be interrupted and put into sleep mode for a short interval of
time to maintain the defined limit. When and how such interruptions occur is defined by a
proprietary algorithm used by EDB Resource Manager.

5.2.1 Setting the CPU Rate Limit for a Resource Group

The ALTER RESOURCE GROUP command with the SET cpu_rate_limit clause is
used to set the CPU rate limit for a resource group.

In the following example the CPU usage limit is set to 50% for resgrp_a, 40% for
resgrp_b and 30% for resgrp_c. This means that the combined CPU usage of all
processes assigned to resgrp_a is maintained at approximately 50%. Similarly, for all
processes in resgrp_b, the combined CPU usage is kept to approximately 40%, etc.

edb=# ALTER RESOURCE GROUP resgrp_a SET cpu_rate_limit TO .5;

ALTER RESOURCE GROUP
edb=# ALTER RESOURCE GROUP resgrp_b SET cpu_rate_limit TO .4;
ALTER RESOURCE GROUP
edb=# ALTER RESOURCE GROUP resgrp_c SET cpu_rate_limit TO .3;
ALTER RESOURCE GROUP

The following query shows the settings of cpu_rate_limit in the catalog.

edb=# SELECT rgrpname, rgrpcpuratelimit FROM edb_resource_group;

rgrpname | rgrpcpuratelimit
----------+------------------
resgrp_a | 0.5
resgrp_b | 0.4

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 205

 EDB Postgres Advanced Server Guide

resgrp_c | 0.3
(3 rows)

Changing the cpu_rate_limit of a resource group not only affects new processes that
are assigned to the group, but any currently running processes that are members of the
group are immediately affected by the change. That is, if the cpu_rate_limit is
changed from .5 to .3, currently running processes in the group would be throttled
downward so that the aggregate group CPU usage would be near 30% instead of 50%.

To illustrate the effect of setting the CPU rate limit for resource groups, the following
examples use a CPU-intensive calculation of 20000 factorial (multiplication of 20000 *
19999 * 19998, etc.) performed by the query SELECT 20000!; run in the psql
command line utility.

The resource groups with the CPU rate limit settings shown in the previous query are
used in these examples.

5.2.2 Example Single Process in a Single Group

The following shows that the current process is set to use resource group resgrp_b. The
factorial calculation is then started.

edb=# SET edb_resource_group TO resgrp_b;

SET
edb=# SHOW edb_resource_group;
edb_resource_group
--------------------
resgrp_b
(1 row)
edb=# SELECT 20000!;

In a second session, the Linux top command is used to display the CPU usage as shown
under the %CPU column. The following is a snapshot at an arbitrary point in time as the
top command output periodically changes.

$ top
top - 16:37:03 up 4:15, 7 users, load average: 0.49, 0.20, 0.38
Tasks: 202 total, 1 running, 201 sleeping, 0 stopped, 0 zombie
Cpu(s): 42.7%us, 2.3%sy, 0.0%ni, 55.0%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0
Mem: 1025624k total, 791160k used, 234464k free, 23400k buffers
Swap: 103420k total, 13404k used, 90016k free, 373504k cached

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND

28915 enterpri 20 0 195m 5900 4212 S 39.9 0.6 3:36.98 edb-postgres
1033 root 20 0 171m 77m 2960 S 1.0 7.8 3:43.96 Xorg
3040 user 20 0 278m 22m 14m S 1.0 2.2 3:41.72 knotify4
.
.
.

The psql session performing the factorial calculation is shown by the row where edb-
postgres appears under the COMMAND column. The CPU usage of the session shown
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 206
 EDB Postgres Advanced Server Guide

under the %CPU column shows 39.9, which is close to the 40% CPU limit set for resource
group resgrp_b.

By contrast, if the psql session is removed from the resource group and the factorial
calculation is performed again, the CPU usage is much higher.

edb=# SET edb_resource_group TO DEFAULT;

SET
edb=# SHOW edb_resource_group;
edb_resource_group
--------------------

(1 row)

edb=# SELECT 20000!;

Under the %CPU column for edb-postgres, the CPU usage is now 93.6, which is
significantly higher than the 39.9 when the process was part of the resource group.

$ top
top - 16:43:03 up 4:21, 7 users, load average: 0.66, 0.33, 0.37
Tasks: 202 total, 5 running, 197 sleeping, 0 stopped, 0 zombie
Cpu(s): 96.7%us, 3.3%sy, 0.0%ni, 0.0%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0
Mem: 1025624k total, 791228k used, 234396k free, 23560k buffers
Swap: 103420k total, 13404k used, 90016k free, 373508k cached

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND

28915 enterpri 20 0 195m 5900 4212 R 93.6 0.6 5:01.56 edb-postgres
1033 root 20 0 171m 77m 2960 S 1.0 7.8 3:48.15 Xorg
2907 user 20 0 98.7m 11m 9100 S 0.3 1.2 0:46.51 vmware-user-lo
.
.
.

5.2.3 Example Multiple Processes in a Single Group

As stated previously, the CPU rate limit applies to the aggregate of all processes in the
resource group. This concept is illustrated in the following example.

The factorial calculation is performed simultaneously in two separate psql sessions,

each of which has been added to resource group resgrp_b that has cpu_rate_limit
set to .4 (CPU usage of 40%).

Session 1:

edb=# SET edb_resource_group TO resgrp_b;

SET
edb=# SHOW edb_resource_group;
edb_resource_group
--------------------
resgrp_b
(1 row)

edb=# SELECT 20000!;

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 207

 EDB Postgres Advanced Server Guide

Session 2:

edb=# SET edb_resource_group TO resgrp_b;

SET
edb=# SHOW edb_resource_group;
edb_resource_group
--------------------
resgrp_b
(1 row)

edb=# SELECT 20000!;

A third session monitors the CPU usage.

$ top
top - 16:53:03 up 4:31, 7 users, load average: 0.31, 0.19, 0.27
Tasks: 202 total, 1 running, 201 sleeping, 0 stopped, 0 zombie
Cpu(s): 41.2%us, 3.0%sy, 0.0%ni, 55.8%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0
Mem: 1025624k total, 792020k used, 233604k free, 23844k buffers
Swap: 103420k total, 13404k used, 90016k free, 373508k cached

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND

29857 enterpri 20 0 195m 4708 3312 S 19.9 0.5 0:57.35 edb-postgres
28915 enterpri 20 0 195m 5900 4212 S 19.6 0.6 5:35.49 edb-postgres
3040 user 20 0 278m 22m 14m S 1.0 2.2 3:54.99 knotify4
1033 root 20 0 171m 78m 2960 S 0.3 7.8 3:55.71 Xorg
.
.
.

There are now two processes named edb-postgres with %CPU values of 19.9 and 19.6,
whose sum is close to the 40% CPU usage set for resource group resgrp_b.

The following command sequence displays the sum of all edb-postgres processes
sampled over half second time intervals. This shows how the total CPU usage of the
processes in the resource group changes over time as EDB Resource Manager throttles
the processes to keep the total resource group CPU usage near 40%.

$ while [[ 1 -eq 1 ]]; do top -d0.5 -b -n2 | grep edb-postgres | awk '{ SUM
+= $9} END { print SUM / 2 }'; done
37.2
39.1
38.9
38.3
44.7
39.2
42.5
39.1
39.2
39.2
41
42.85
46.1
.
.
.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 208

 EDB Postgres Advanced Server Guide

5.2.4 Example Multiple Processes in Multiple Groups

In this example, two additional psql sessions are used along with the previous two
sessions. The third and fourth sessions perform the same factorial calculation within
resource group resgrp_c with a cpu_rate_limit of .3 (30% CPU usage).

Session 3:

edb=# SET edb_resource_group TO resgrp_c;

SET
edb=# SHOW edb_resource_group;
edb_resource_group
--------------------
resgrp_c
(1 row)

edb=# SELECT 20000!;

Session 4:

edb=# SET edb_resource_group TO resgrp_c;

SET
edb=# SHOW edb_resource_group;
edb_resource_group
--------------------
resgrp_c
(1 row)

edb=# SELECT 20000!;

The top command displays the following output.

$ top
top - 17:45:09 up 5:23, 8 users, load average: 0.47, 0.17, 0.26
Tasks: 203 total, 4 running, 199 sleeping, 0 stopped, 0 zombie
Cpu(s): 70.2%us, 0.0%sy, 0.0%ni, 29.8%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0
Mem: 1025624k total, 806140k used, 219484k free, 25296k buffers
Swap: 103420k total, 13404k used, 90016k free, 374092k cached

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND

29857 enterpri 20 0 195m 4820 3324 S 19.9 0.5 4:25.02 edb-postgres
28915 enterpri 20 0 195m 5900 4212 R 19.6 0.6 9:07.50 edb-postgres
29023 enterpri 20 0 195m 4744 3248 R 16.3 0.5 4:01.73 edb-postgres
11019 enterpri 20 0 195m 4120 2764 R 15.3 0.4 0:04.92 edb-postgres
2907 user 20 0 98.7m 12m 9112 S 1.3 1.2 0:56.54 vmware-user-lo
3040 user 20 0 278m 22m 14m S 1.3 2.2 4:38.73 knotify4

The two resource groups in use have CPU usage limits of 40% and 30%. The sum of the
%CPU column for the first two edb-postgres processes is 39.5 (approximately 40%,
which is the limit for resgrp_b) and the sum of the %CPU column for the third and
fourth edb-postgres processes is 31.6 (approximately 30%, which is the limit for
resgrp_c).

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 209

 EDB Postgres Advanced Server Guide

The sum of the CPU usage limits of the two resource groups to which these processes
belong is 70%. The following output shows that the sum of the four processes borders
around 70%.

$ while [[ 1 -eq 1 ]]; do top -d0.5 -b -n2 | grep edb-postgres | awk '{ SUM
+= $9} END { print SUM / 2 }'; done
61.8
76.4
72.6
69.55
64.55
79.95
68.55
71.25
74.85
62
74.85
76.9
72.4
65.9
74.9
68.25

By contrast, if three sessions are processing where two sessions remain in resgrp_b, but
the third session does not belong to any resource group, the top command shows the
following output.

$ top
top - 17:24:55 up 5:03, 7 users, load average: 1.00, 0.41, 0.38
Tasks: 199 total, 3 running, 196 sleeping, 0 stopped, 0 zombie
Cpu(s): 99.7%us, 0.3%sy, 0.0%ni, 0.0%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0
Mem: 1025624k total, 797692k used, 227932k free, 24724k buffers
Swap: 103420k total, 13404k used, 90016k free, 374068k cached

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND

29023 enterpri 20 0 195m 4744 3248 R 58.6 0.5 2:53.75 edb-postgres
28915 enterpri 20 0 195m 5900 4212 S 18.9 0.6 7:58.45 edb-postgres
29857 enterpri 20 0 195m 4820 3324 S 18.9 0.5 3:14.85 edb-postgres
1033 root 20 0 174m 81m 2960 S 1.7 8.2 4:26.50 Xorg
3040 user 20 0 278m 22m 14m S 1.0 2.2 4:21.20 knotify4

The second and third edb-postgres processes belonging to the resource group where
the CPU usage is limited to 40%, have a total CPU usage of 37.8. However, the first
edb-postgres process has a 58.6% CPU usage as it is not within a resource group, and
basically utilizes the remaining, available CPU resources on the system.

Likewise, the following output shows the sum of all three sessions is around 95% since
one of the sessions has no set limit on its CPU usage.

$ while [[ 1 -eq 1 ]]; do top -d0.5 -b -n2 | grep edb-postgres | awk '{ SUM
+= $9} END { print SUM / 2 }'; done
96
90.35
92.55
96.4
94.1
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 210
 EDB Postgres Advanced Server Guide

90.7
95.7
95.45
93.65
87.95
96.75
94.25
95.45
97.35
92.9
96.05
96.25
94.95
.
.
.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 211

 EDB Postgres Advanced Server Guide

5.3 Dirty Buffer Throttling

Writing to shared buffers is controlled by setting the dirty_rate_limit resource type

parameter.

Set the dirty_rate_limit parameter to the number of kilobytes per second for the
combined rate at which all the processes in the group should write to or dirty the shared
buffers. An example setting would be 3072 kilobytes per seconds.

The valid range of the dirty_rate_limit parameter is 0 to 1.67772e+07. A setting of

0 means no dirty rate limit has been set for the resource group.

EDB Resource Manager utilizes dirty buffer throttling to keep the aggregate, shared
buffer writing rate of all processes in the group near the limit specified by the
dirty_rate_limit parameter. A process in the group may be interrupted and put into
sleep mode for a short interval of time to maintain the defined limit. When and how such
interruptions occur is defined by a proprietary algorithm used by EDB Resource
Manager.

5.3.1 Setting the Dirty Rate Limit for a Resource Group

The ALTER RESOURCE GROUP command with the SET dirty_rate_limit clause is
used to set the dirty rate limit for a resource group.

In the following example the dirty rate limit is set to 12288 kilobytes per second for
resgrp_a, 6144 kilobytes per second for resgrp_b and 3072 kilobytes per second for
resgrp_c. This means that the combined writing rate to the shared buffer of all
processes assigned to resgrp_a is maintained at approximately 12288 kilobytes per
second. Similarly, for all processes in resgrp_b, the combined writing rate to the shared
buffer is kept to approximately 6144 kilobytes per second, etc.

edb=# ALTER RESOURCE GROUP resgrp_a SET dirty_rate_limit TO 12288;

ALTER RESOURCE GROUP
edb=# ALTER RESOURCE GROUP resgrp_b SET dirty_rate_limit TO 6144;
ALTER RESOURCE GROUP
edb=# ALTER RESOURCE GROUP resgrp_c SET dirty_rate_limit TO 3072;
ALTER RESOURCE GROUP

The following query shows the settings of dirty_rate_limit in the catalog.

edb=# SELECT rgrpname, rgrpdirtyratelimit FROM edb_resource_group;

rgrpname | rgrpdirtyratelimit
----------+--------------------
resgrp_a | 12288
resgrp_b | 6144
resgrp_c | 3072
(3 rows)
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 212
 EDB Postgres Advanced Server Guide

Changing the dirty_rate_limit of a resource group not only affects new processes
that are assigned to the group, but any currently running processes that are members of
the group are immediately affected by the change. That is, if the dirty_rate_limit is
changed from 12288 to 3072, currently running processes in the group would be throttled
downward so that the aggregate group dirty rate would be near 3072 kilobytes per second
instead of 12288 kilobytes per second.

To illustrate the effect of setting the dirty rate limit for resource groups, the following
examples use the following table for intensive I/O operations.

CREATE TABLE t1 (c1 INTEGER, c2 CHARACTER(500)) WITH (FILLFACTOR = 10);

The FILLFACTOR = 10 clause results in INSERT commands packing rows up to only

10% per page. This results in a larger sampling of dirty shared blocks for the purpose of
these examples.

The pg_stat_statements module is used to display the number of shared buffer

blocks that are dirtied by a SQL command and the amount of time the command took to
execute. This provides the information to calculate the actual kilobytes per second
writing rate for the SQL command, and thus compare it to the dirty rate limit set for a
resource group.

In order to use the pg_stat_statements module, perform the following steps.

Step 1: In the postgresql.conf file, add $libdir/pg_stat_statements to the

shared_preload_libraries configuration parameter as shown by the following.

shared_preload_libraries = '$libdir/dbms_pipe,$libdir/edb_gen,$libdir/pg_stat_statements'

Step 2: Restart the database server.

Step 3: Use the CREATE EXTENSION command to complete the creation of the
pg_stat_statements module.

edb=# CREATE EXTENSION pg_stat_statements SCHEMA public;

CREATE EXTENSION

The pg_stat_statements_reset() function is used to clear out the

pg_stat_statements view for clarity of each example.

The resource groups with the dirty rate limit settings shown in the previous query are
used in these examples.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 213

 EDB Postgres Advanced Server Guide

5.3.2 Example Single Process in a Single Group

The following sequence of commands shows the creation of table t1. The current process
is set to use resource group resgrp_b. The pg_stat_statements view is cleared out
by running the pg_stat_statements_reset() function.

Finally, the INSERT command generates a series of integers from 1 to 10,000 to populate
the table, and dirty approximately 10,000 blocks.

edb=# CREATE TABLE t1 (c1 INTEGER, c2 CHARACTER(500)) WITH (FILLFACTOR = 10);

CREATE TABLE
edb=# SET edb_resource_group TO resgrp_b;
SET
edb=# SHOW edb_resource_group;
edb_resource_group
--------------------
resgrp_b
(1 row)

edb=# SELECT pg_stat_statements_reset();

pg_stat_statements_reset
--------------------------

(1 row)

edb=# INSERT INTO t1 VALUES (generate_series (1,10000), 'aaa');

INSERT 0 10000

The following shows the results from the INSERT command.

edb=# SELECT query, rows, total_time, shared_blks_dirtied FROM

pg_stat_statements;
-[ RECORD 1 ]-------+--------------------------------------------------
query | INSERT INTO t1 VALUES (generate_series (?,?), ?);
rows | 10000
total_time | 13496.184
shared_blks_dirtied | 10003

The actual dirty rate is calculated as follows.

The number of blocks dirtied per millisecond (ms) is 10003 blocks / 13496.184
ms, which yields 0.74117247 blocks per millisecond.
Multiply the result by 1000 to give the number of shared blocks dirtied per second
(1 second = 1000 ms), which yields 741.17247 blocks per second.
Multiply the result by 8.192 to give the number of kilobytes dirtied per second (1
block = 8.192 kilobytes), which yields approximately 6072 kilobytes per second.

Note that the actual dirty rate of 6072 kilobytes per second is close to the dirty rate limit
for the resource group, which is 6144 kilobytes per second.

By contrast, if the steps are repeated again without the process belonging to any resource
group, the dirty buffer rate is much higher.
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 214
 EDB Postgres Advanced Server Guide

edb=# CREATE TABLE t1 (c1 INTEGER, c2 CHARACTER(500)) WITH (FILLFACTOR = 10);

CREATE TABLE
edb=# SHOW edb_resource_group;
edb_resource_group
--------------------

(1 row)

edb=# SELECT pg_stat_statements_reset();

pg_stat_statements_reset
--------------------------

(1 row)

edb=# INSERT INTO t1 VALUES (generate_series (1,10000), 'aaa');

INSERT 0 10000

The following shows the results from the INSERT command without the usage of a
resource group.

edb=# SELECT query, rows, total_time, shared_blks_dirtied FROM

pg_stat_statements;
-[ RECORD 1 ]-------+--------------------------------------------------
query | INSERT INTO t1 VALUES (generate_series (?,?), ?);
rows | 10000
total_time | 2432.165
shared_blks_dirtied | 10003

First, note the total time was only 2432.165 milliseconds as compared to 13496.184
milliseconds when a resource group with a dirty rate limit set to 6144 kilobytes per
second was used.

The actual dirty rate without the use of a resource group is calculated as follows.

The number of blocks dirtied per millisecond (ms) is 10003 blocks / 2432.165 ms,
which yields 4.112797 blocks per millisecond.
Multiply the result by 1000 to give the number of shared blocks dirtied per second
(1 second = 1000 ms), which yields 4112.797 blocks per second.
Multiply the result by 8.192 to give the number of kilobytes dirtied per second (1
block = 8.192 kilobytes), which yields approximately 33692 kilobytes per second.

Note that the actual dirty rate of 33692 kilobytes per second is significantly higher than
when the resource group with a dirty rate limit of 6144 kilobytes per second was used.

5.3.3 Example Multiple Processes in a Single Group

As stated previously, the dirty rate limit applies to the aggregate of all processes in the
resource group. This concept is illustrated in the following example.

For this example the inserts are performed simultaneously on two different tables in two
separate psql sessions, each of which has been added to resource group resgrp_b that
has a dirty_rate_limit set to 6144 kilobytes per second.
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 215
 EDB Postgres Advanced Server Guide

Session 1:

edb=# CREATE TABLE t1 (c1 INTEGER, c2 CHARACTER(500)) WITH (FILLFACTOR = 10);

CREATE TABLE
edb=# SET edb_resource_group TO resgrp_b;
SET
edb=# SHOW edb_resource_group;
edb_resource_group
--------------------
resgrp_b
(1 row)

edb=# INSERT INTO t1 VALUES (generate_series (1,10000), 'aaa');

INSERT 0 10000

Session 2:

edb=# CREATE TABLE t2 (c1 INTEGER, c2 CHARACTER(500)) WITH (FILLFACTOR = 10);

CREATE TABLE
edb=# SET edb_resource_group TO resgrp_b;
SET
edb=# SHOW edb_resource_group;
edb_resource_group
--------------------
resgrp_b
(1 row)

edb=# SELECT pg_stat_statements_reset();

pg_stat_statements_reset
--------------------------
(1 row)

edb=# INSERT INTO t2 VALUES (generate_series (1,10000), 'aaa');

INSERT 0 10000

Note: The INSERT commands in session 1 and session 2 were started after the SELECT
pg_stat_statements_reset() command in session 2 was run.

The following shows the results from the INSERT commands in the two sessions.
RECORD 3 shows the results from session 1. RECORD 2 shows the results from session 2.

edb=# SELECT query, rows, total_time, shared_blks_dirtied FROM

pg_stat_statements;
-[ RECORD 1 ]-------+--------------------------------------------------
query | SELECT pg_stat_statements_reset();
rows | 1
total_time | 0.43
shared_blks_dirtied | 0
-[ RECORD 2 ]-------+--------------------------------------------------
query | INSERT INTO t2 VALUES (generate_series (?,?), ?);
rows | 10000
total_time | 30591.551
shared_blks_dirtied | 10003
-[ RECORD 3 ]-------+--------------------------------------------------
query | INSERT INTO t1 VALUES (generate_series (?,?), ?);
rows | 10000
total_time | 33215.334
shared_blks_dirtied | 10003
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 216
 EDB Postgres Advanced Server Guide

First, note the total time was 33215.334 milliseconds for session 1 and 30591.551
milliseconds for session 2. When only one session was active in the same resource group
as shown in the first example, the time was 13496.184 milliseconds. Thus more active
processes in the resource group result in a slower dirty rate for each active process in the
group. This is shown in the following calculations.

The actual dirty rate for session 1 is calculated as follows.

The number of blocks dirtied per millisecond (ms) is 10003 blocks / 33215.334
ms, which yields 0.30115609 blocks per millisecond.
Multiply the result by 1000 to give the number of shared blocks dirtied per second
(1 second = 1000 ms), which yields 301.15609 blocks per second.
Multiply the result by 8.192 to give the number of kilobytes dirtied per second (1
block = 8.192 kilobytes), which yields approximately 2467 kilobytes per second.

The actual dirty rate for session 2 is calculated as follows.

The number of blocks dirtied per millisecond (ms) is 10003 blocks / 30591.551
ms, which yields 0.32698571 blocks per millisecond.
Multiply the result by 1000 to give the number of shared blocks dirtied per second
(1 second = 1000 ms), which yields 326.98571 blocks per second.
Multiply the result by 8.192 to give the number of kilobytes dirtied per second (1
block = 8.192 kilobytes), which yields approximately 2679 kilobytes per second.

The combined dirty rate from session 1 (2467 kilobytes per second) and from session 2
(2679 kilobytes per second) yields 5146 kilobytes per second, which is below the set
dirty rate limit of the resource group (6144 kilobytes per seconds).

5.3.4 Example Multiple Processes in Multiple Groups

In this example, two additional psql sessions are used along with the previous two
sessions. The third and fourth sessions perform the same INSERT command in resource
group resgrp_c with a dirty_rate_limit of 3072 kilobytes per second.

Sessions 1 and 2 are repeated as illustrated in the prior example using resource group
resgrp_b. with a dirty_rate_limit of 6144 kilobytes per second.

Session 3:

edb=# CREATE TABLE t3 (c1 INTEGER, c2 CHARACTER(500)) WITH (FILLFACTOR = 10);

CREATE TABLE
edb=# SET edb_resource_group TO resgrp_c;
SET
edb=# SHOW edb_resource_group;
edb_resource_group
--------------------

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 217

 EDB Postgres Advanced Server Guide

resgrp_c
(1 row)

edb=# INSERT INTO t3 VALUES (generate_series (1,10000), 'aaa');

INSERT 0 10000

Session 4:

edb=# CREATE TABLE t4 (c1 INTEGER, c2 CHARACTER(500)) WITH (FILLFACTOR = 10);

CREATE TABLE
edb=# SET edb_resource_group TO resgrp_c;
SET
edb=# SHOW edb_resource_group;
edb_resource_group
--------------------
resgrp_c
(1 row)

edb=# SELECT pg_stat_statements_reset();

pg_stat_statements_reset
--------------------------

(1 row)

edb=# INSERT INTO t4 VALUES (generate_series (1,10000), 'aaa');

INSERT 0 10000

Note: The INSERT commands in all four sessions were started after the SELECT
pg_stat_statements_reset() command in session 4 was run.

The following shows the results from the INSERT commands in the four sessions.
RECORD 3 shows the results from session 1. RECORD 2 shows the results from session 2.
RECORD 4 shows the results from session 3. RECORD 5 shows the results from session 4.

edb=# SELECT query, rows, total_time, shared_blks_dirtied FROM

pg_stat_statements;
-[ RECORD 1 ]-------+--------------------------------------------------
query | SELECT pg_stat_statements_reset();
rows | 1
total_time | 0.467
shared_blks_dirtied | 0
-[ RECORD 2 ]-------+--------------------------------------------------
query | INSERT INTO t2 VALUES (generate_series (?,?), ?);
rows | 10000
total_time | 31343.458
shared_blks_dirtied | 10003
-[ RECORD 3 ]-------+--------------------------------------------------
query | INSERT INTO t1 VALUES (generate_series (?,?), ?);
rows | 10000
total_time | 28407.435
shared_blks_dirtied | 10003
-[ RECORD 4 ]-------+--------------------------------------------------
query | INSERT INTO t3 VALUES (generate_series (?,?), ?);
rows | 10000
total_time | 52727.846
shared_blks_dirtied | 10003
-[ RECORD 5 ]-------+--------------------------------------------------
query | INSERT INTO t4 VALUES (generate_series (?,?), ?);

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 218

 EDB Postgres Advanced Server Guide

rows | 10000
total_time | 56063.697
shared_blks_dirtied | 10003

First note that the times of session 1 (28407.435) and session 2 (31343.458) are close to
each other as they are both in the same resource group with dirty_rate_limit set to
6144, as compared to the times of session 3 (52727.846) and session 4 (56063.697),
which are in the resource group with dirty_rate_limit set to 3072. The latter group
has a slower dirty rate limit so the expected processing time is longer as is the case for
sessions 3 and 4.

The actual dirty rate for session 1 is calculated as follows.

The number of blocks dirtied per millisecond (ms) is 10003 blocks / 28407.435
ms, which yields 0.35212612 blocks per millisecond.
Multiply the result by 1000 to give the number of shared blocks dirtied per second
(1 second = 1000 ms), which yields 352.12612 blocks per second.
Multiply the result by 8.192 to give the number of kilobytes dirtied per second (1
block = 8.192 kilobytes), which yields approximately 2885 kilobytes per second.

The actual dirty rate for session 2 is calculated as follows.

The number of blocks dirtied per millisecond (ms) is 10003 blocks / 31343.458
ms, which yields 0.31914156 blocks per millisecond.
Multiply the result by 1000 to give the number of shared blocks dirtied per second
(1 second = 1000 ms), which yields 319.14156 blocks per second.
Multiply the result by 8.192 to give the number of kilobytes dirtied per second (1
block = 8.192 kilobytes), which yields approximately 2614 kilobytes per second.

The combined dirty rate from session 1 (2885 kilobytes per second) and from session 2
(2614 kilobytes per second) yields 5499 kilobytes per second, which is near the set dirty
rate limit of the resource group (6144 kilobytes per seconds).

The actual dirty rate for session 3 is calculated as follows.

The number of blocks dirtied per millisecond (ms) is 10003 blocks / 52727.846
ms, which yields 0.18971001 blocks per millisecond.
Multiply the result by 1000 to give the number of shared blocks dirtied per second
(1 second = 1000 ms), which yields 189.71001 blocks per second.
Multiply the result by 8.192 to give the number of kilobytes dirtied per second (1
block = 8.192 kilobytes), which yields approximately 1554 kilobytes per second.

The actual dirty rate for session 4 is calculated as follows.

The number of blocks dirtied per millisecond (ms) is 10003 blocks / 56063.697
ms, which yields 0.17842205 blocks per millisecond.
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 219
 EDB Postgres Advanced Server Guide

Multiply the result by 1000 to give the number of shared blocks dirtied per second
(1 second = 1000 ms), which yields 178.42205 blocks per second.
Multiply the result by 8.192 to give the number of kilobytes dirtied per second (1
block = 8.192 kilobytes), which yields approximately 1462 kilobytes per second.

The combined dirty rate from session 3 (1554 kilobytes per second) and from session 4
(1462 kilobytes per second) yields 3016 kilobytes per second, which is near the set dirty
rate limit of the resource group (3072 kilobytes per seconds).

Thus, this demonstrates how EDB Resource Manager keeps the aggregate dirty rate of
the active processes in its groups close to the dirty rate limit set for each group.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 220

 EDB Postgres Advanced Server Guide

5.4 System Catalogs

This section describes the system catalogs that store the resource group information used
by EDB Resource Manager.

5.4.1 edb_all_resource_groups

The following table lists the information available in the edb_all_resource_groups

catalog:

Column Type Description

group_name name The name of the resource group.
active_processes integer Number of currently active processes in the
resource group.
cpu_rate_limit float8 Maximum CPU rate limit for the resource
group. 0 means no limit.
per_process_cpu_rate_limit float8 Maximum CPU rate limit per currently active
process in the resource group.
dirty_rate_limit float8 Maximum dirty rate limit for a resource
group. 0 means no limit.
per_process_dirty_rate_limit float8 Maximum dirty rate limit per currently active
process in the resource group.

5.4.2 edb_resource_group

The following table lists the information available in the edb_resource_group

catalog:

Column Type Description

rgrpname name The name of the resource group.
rgrpcpuratelimit float8 Maximum CPU rate limit for a resource
group. 0 means no limit.
rgrpdirtyratelimit float8 Maximum dirty rate limit for a resource group.
0 means no limit.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 221

 EDB Postgres Advanced Server Guide

6 libpq C Library
libpq is the C application programmers interface to Advanced Server. libpq is a set of
library functions that allow client programs to pass queries to the Advanced Server and to
receive the results of these queries.

libpq is also the underlying engine for several other EnterpriseDB application interfaces
including those written for C++, Perl, Python, Tcl and ECPG. So some aspects of libpqs
behavior will be important to the user if one of those packages is used.

Client programs that use libpq must include the header file libpq-fe.h and must link
with the libpq library.

6.1.1 Using libpq with EnterpriseDB SPL

The EnterpriseDB SPL language can be used with the libpq interface library, providing
support for:

Procedures, functions, packages

Prepared statements
REFCURSORs
Static cursors
structs and typedefs
Arrays
DML and DDL operations
IN/OUT/IN OUT parameters

6.1.2 REFCURSOR Support

In earlier releases, Advanced Server provided support for REFCURSORs through the
following libpq functions; these functions should now be considered deprecated:

PQCursorResult()
PQgetCursorResult()
PQnCursor()

You may now use PQexec() and PQgetvalue() to retrieve a REFCURSOR returned by
an SPL (or PL/pgSQL) function. A REFCURSOR is returned in the form of a null-
terminated string indicating the name of the cursor. Once you have the name of the
cursor, you can execute one or more FETCH statements to retrieve the values exposed
through the cursor.

Please note that the samples that follow do not include error-handling code that would be
required in a real-world client application.
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 222
 EDB Postgres Advanced Server Guide

Returning a Single REFCURSOR

The following example shows an SPL function that returns a value of type REFCURSOR:

CREATE OR REPLACE FUNCTION getEmployees(p_deptno NUMERIC)

RETURN REFCURSOR AS
result REFCURSOR;
BEGIN
OPEN result FOR SELECT * FROM emp WHERE deptno = p_deptno;

RETURN result;
END;

This function expects a single parameter, p_deptno, and returns a REFCURSOR that
holds the result set for the SELECT query shown in the OPEN statement. The OPEN
statement executes the query and stores the result set in a cursor. The server constructs a
name for that cursor and stores the name in a variable (named result). The function
then returns the name of the cursor to the caller.

To call this function from a C client using libpq, you can use PQexec() and
PQgetvalue():

#include <stdio.h>
#include <stdlib.h>
#include "libpq-fe.h"

static void fetchAllRows(PGconn *conn,

const char *cursorName,
const char *description);
static void fail(PGconn *conn, const char *msg);

int
main(int argc, char *argv[])
{
PGconn *conn = PQconnectdb(argv[1]);
PGresult *result;

if (PQstatus(conn) != CONNECTION_OK)
fail(conn, PQerrorMessage(conn));

result = PQexec(conn, "BEGIN TRANSACTION");

if (PQresultStatus(result) != PGRES_COMMAND_OK)
fail(conn, PQerrorMessage(conn));

PQclear(result);

result = PQexec(conn, "SELECT * FROM getEmployees(10)");

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 223

 EDB Postgres Advanced Server Guide

if (PQresultStatus(result) != PGRES_TUPLES_OK)
fail(conn, PQerrorMessage(conn));

fetchAllRows(conn, PQgetvalue(result, 0, 0), "employees");

PQclear(result);

PQexec(conn, "COMMIT");

PQfinish(conn);

exit(0);
}

static void
fetchAllRows(PGconn *conn,
const char *cursorName,
const char *description)
{
size_t commandLength = strlen("FETCH ALL FROM ") +
strlen(cursorName) + 3;

char *commandText = malloc(commandLength);

PGresult *result;
int row;

sprintf(commandText, "FETCH ALL FROM \"%s\"", cursorName);

result = PQexec(conn, commandText);

if (PQresultStatus(result) != PGRES_TUPLES_OK)
fail(conn, PQerrorMessage(conn));

printf("-- %s --\n", description);

for (row = 0; row < PQntuples(result); row++)

{
const char *delimiter = "\t";
int col;

for (col = 0; col < PQnfields(result); col++)

{
printf("%s%s", delimiter, PQgetvalue(result, row, col));
delimiter = ",";
}

printf("\n");
}

PQclear(result);
free(commandText);
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 224
 EDB Postgres Advanced Server Guide

static void
fail(PGconn *conn, const char *msg)
{
fprintf(stderr, "%s\n", msg);

if (conn != NULL)
PQfinish(conn);

exit(-1);
}

The code sample contains a line of code that calls the getEmployees() function, and
returns a result set that contains all of the employees in department 10:

result = PQexec(conn, "SELECT * FROM getEmployees(10)");

The PQexec() function returns a result set handle to the C program. The result set will
contain exactly one value; that value is the name of the cursor as returned by
getEmployees().

Once you have the name of the cursor, you can use the SQL FETCH statement to retrieve
the rows in that cursor. The function fetchAllRows() builds a FETCH ALL statement,
executes that statement, and then prints the result set of the FETCH ALL statement.

The output of this program is shown below:

-- employees --
7782,CLARK,MANAGER,7839,09-JUN-81 00:00:00,2450.00,,10
7839,KING,PRESIDENT,,17-NOV-81 00:00:00,5000.00,,10
7934,MILLER,CLERK,7782,23-JAN-82 00:00:00,1300.00,,10

Returning Multiple REFCURSORs

The next example returns two REFCURSORs:

The first REFCURSOR contains the name of a cursor (employees) that contains
all employees who work in a department within the range specified by the caller.

The second REFCURSOR contains the name of a cursor (departments) that

contains all of the departments in the range specified by the caller.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 225

 EDB Postgres Advanced Server Guide

In this example, instead of returning a single REFCURSOR, the function returns a SETOF
REFCURSOR (which means 0 or more REFCURSORS). One other important difference is
that the libpq program should not expect a single REFCURSOR in the result set, but should
expect two rows, each of which will contain a single value (the first row contains the
name of the employees cursor, and the second row contains the name of the
departments cursor).

CREATE OR REPLACE FUNCTION getEmpsAndDepts(p_min NUMERIC,

p_max NUMERIC)
RETURN SETOF REFCURSOR AS
employees REFCURSOR;
departments REFCURSOR;
BEGIN
OPEN employees FOR
SELECT * FROM emp WHERE deptno BETWEEN p_min AND p_max;
RETURN NEXT employees;

OPEN departments FOR

SELECT * FROM dept WHERE deptno BETWEEN p_min AND p_max;
RETURN NEXT departments;
END;

As in the previous example, you can use PQexec() and PQgetvalue() to call the SPL
function:

#include <stdio.h>
#include <stdlib.h>
#include "libpq-fe.h"

static void fetchAllRows(PGconn *conn,

const char *cursorName,
const char *description);
static void fail(PGconn *conn, const char *msg);

int
main(int argc, char *argv[])
{
PGconn *conn = PQconnectdb(argv[1]);
PGresult *result;

if (PQstatus(conn) != CONNECTION_OK)
fail(conn, PQerrorMessage(conn));

result = PQexec(conn, "BEGIN TRANSACTION");

if (PQresultStatus(result) != PGRES_COMMAND_OK)
fail(conn, PQerrorMessage(conn));

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 226

 EDB Postgres Advanced Server Guide

PQclear(result);

result = PQexec(conn, "SELECT * FROM getEmpsAndDepts(20, 30)");

if (PQresultStatus(result) != PGRES_TUPLES_OK)
fail(conn, PQerrorMessage(conn));

fetchAllRows(conn, PQgetvalue(result, 0, 0), "employees");

fetchAllRows(conn, PQgetvalue(result, 1, 0), "departments");

PQclear(result);

PQexec(conn, "COMMIT");

PQfinish(conn);

exit(0);
}

static void
fetchAllRows(PGconn *conn,
const char *cursorName,
const char *description)
{
size_t commandLength = strlen("FETCH ALL FROM ") +
strlen(cursorName) + 3;
char *commandText = malloc(commandLength);
PGresult *result;
int row;

sprintf(commandText, "FETCH ALL FROM \"%s\"", cursorName);

result = PQexec(conn, commandText);

if (PQresultStatus(result) != PGRES_TUPLES_OK)
fail(conn, PQerrorMessage(conn));

printf("-- %s --\n", description);

for (row = 0; row < PQntuples(result); row++)

{
const char *delimiter = "\t";
int col;

for (col = 0; col < PQnfields(result); col++)

{
printf("%s%s", delimiter, PQgetvalue(result, row, col));
delimiter = ",";
}

printf("\n");
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 227
 EDB Postgres Advanced Server Guide

PQclear(result);
free(commandText);
}

static void
fail(PGconn *conn, const char *msg)
{
fprintf(stderr, "%s\n", msg);

if (conn != NULL)
PQfinish(conn);

exit(-1);
}

If you call getEmpsAndDepts(20, 30), the server will return a cursor that contains all
employees who work in department 20 or 30, and a second cursor containing the
description of departments 20 and 30.

-- employees --
7369,SMITH,CLERK,7902,17-DEC-80 00:00:00,800.00,,20
7499,ALLEN,SALESMAN,7698,20-FEB-81 00:00:00,1600.00,300.00,30
7521,WARD,SALESMAN,7698,22-FEB-81 00:00:00,1250.00,500.00,30
7566,JONES,MANAGER,7839,02-APR-81 00:00:00,2975.00,,20
7654,MARTIN,SALESMAN,7698,28-SEP-81 00:00:00,1250.00,1400.00,30
7698,BLAKE,MANAGER,7839,01-MAY-81 00:00:00,2850.00,,30
7788,SCOTT,ANALYST,7566,19-APR-87 00:00:00,3000.00,,20
7844,TURNER,SALESMAN,7698,08-SEP-81 00:00:00,1500.00,0.00,30
7876,ADAMS,CLERK,7788,23-MAY-87 00:00:00,1100.00,,20
7900,JAMES,CLERK,7698,03-DEC-81 00:00:00,950.00,,30
7902,FORD,ANALYST,7566,03-DEC-81 00:00:00,3000.00,,20
-- departments --
20,RESEARCH,DALLAS
30,SALES,CHICAGO

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 228

 EDB Postgres Advanced Server Guide

6.1.3 Array Binding

Advanced Server's array binding functionality allows you to send an array of data across
the network in a single round-trip. When the back end receives the bulk data, it can use
the data to perform insert or update operations.

Perform bulk operations with a prepared statement; use the following function to prepare
the statement:

PGresult *PQprepare(PGconn *conn,

const char *stmtName,
const char *query,
int nParams,
const Oid *paramTypes);

Details of PQprepare() can be found in the prepared statement section.

The following functions can be used to perform bulk operations:

PQBulkStart
PQexecBulk
PQBulkFinish
PQexecBulkPrepared

6.1.3.1 PQBulkStart

PQBulkStart() initializes bulk operations on the server. You must call this function
before sending bulk data to the server. PQBulkStart() initializes the prepared
statement specified in stmtName to receive data in a format specified by paramFmts.

API Definition

PGresult * PQBulkStart(PGconn *conn,

const char * Stmt_Name,
unsigned int nCol,
const int *paramFmts);

6.1.3.2 PQexecBulk

PQexecBulk() is used to supply data (paramValues) for a statement that was

previously initialized for bulk operation using PQBulkStart().

This function can be used more than once after PQBulkStart() to send multiple blocks
of data. See the example for more details.
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 229
 EDB Postgres Advanced Server Guide

API Definition

PGresult *PQexecBulk(PGconn *conn,

unsigned int nRows,
const char *const * paramValues,
const int *paramLengths);

6.1.3.3 PQBulkFinish
This function completes the current bulk operation. You can use the prepared statement
again without re-preparing it.

API Definition

PGresult *PQBulkFinish(PGconn *conn);

6.1.3.4 PQexecBulkPrepared

Alternatively, you can use the PQexecBulkPrepared() function to perform a bulk

operation with a single function call. PQexecBulkPrepared() sends a request to
execute a prepared statement with the given parameters, and waits for the result. This
function combines the functionality of PQbulkStart(), PQexecBulk(), and
PQBulkFinish(). When using this function, you are not required to initialize or
terminate the bulk operation; this function starts the bulk operation, passes the data to the
server, and terminates the bulk operation.

Specify a previously prepared statement in the place of stmtName. Commands that will
be used repeatedly will be parsed and planned just once, rather than each time they are
executed.

API Definition

PGresult *PQexecBulkPrepared(PGconn *conn,

const char *stmtName,
unsigned int nCols,
unsigned int nRows,
const char *const *paramValues,
const int *paramLengths,
const int *paramFormats);

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 230

 EDB Postgres Advanced Server Guide

6.1.3.5 Example Code (Using PQBulkStart, PQexecBulk,

PQBulkFinish)

The following example uses PGBulkStart, PQexecBulk, and PQBulkFinish.

void InsertDataUsingBulkStyle( PGconn *conn )

{
PGresult *res;
Oid paramTypes[2];
char *paramVals[5][2];
int paramLens[5][2];
int paramFmts[2];
int i;

int a[5] = { 10, 20, 30, 40, 50 };

char b[5][10] = { "Test_1", "Test_2", "Test_3", "Test_4",
"Test_5" };

paramTypes[0] = 23;
paramTypes[1] = 1043;
res = PQprepare( conn, "stmt_1", "INSERT INTO testtable1 values( $1, $2
)", 2, paramTypes );
PQclear( res );

paramFmts[0] = 1; /* Binary format */

paramFmts[1] = 0;

for( i = 0; i < 5; i++ )

{
a[i] = htonl( a[i] );
paramVals[i][0] = &(a[i]);
paramVals[i][1] = b[i];

paramLens[i][0] = 4;
paramLens[i][1] = strlen( b[i] );
}

res = PQBulkStart(conn, "stmt_1", 2, paramFmts);

PQclear( res );
printf( "< -- PQBulkStart -- >\n" );

res = PQexecBulk(conn, 5, (const char *const *)paramVals, (const int

*)paramLens);
PQclear( res );
printf( "< -- PQexecBulk -- >\n" );

res = PQexecBulk(conn, 5, (const char *const *)paramVals, (const int

*)paramLens);
PQclear( res );
printf( "< -- PQexecBulk -- >\n" );

res = PQBulkFinish(conn);
PQclear( res );
printf( "< -- PQBulkFinish -- >\n" );
}

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 231

 EDB Postgres Advanced Server Guide

6.1.3.6 Example Code (Using PQexecBulkPrepared)

The following example uses PQexecBulkPrepared.

void InsertDataUsingBulkStyleCombinedVersion( PGconn *conn )

{
PGresult *res;
Oid paramTypes[2];
char *paramVals[5][2];
int paramLens[5][2];
int paramFmts[2];
int i;

int a[5] = { 10, 20, 30, 40, 50 };

char b[5][10] = { "Test_1", "Test_2", "Test_3", "Test_4",
"Test_5" };

paramTypes[0] = 23;
paramTypes[1] = 1043;
res = PQprepare( conn, "stmt_2", "INSERT INTO testtable1 values( $1, $2
)", 2, paramTypes );
PQclear( res );

paramFmts[0] = 1; /* Binary format */

paramFmts[1] = 0;

for( i = 0; i < 5; i++ )

{
a[i] = htonl( a[i] );
paramVals[i][0] = &(a[i]);
paramVals[i][1] = b[i];

paramLens[i][0] = 4;
paramLens[i][1] = strlen( b[i] );
}

res = PQexecBulkPrepared(conn, "stmt_2", 2, 5, (const char *const

*)paramVals,(const int *)paramLens, (const int *)paramFmts);
PQclear( res );
}

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 232

 EDB Postgres Advanced Server Guide

7 Debugger
The Debugger is a tool that gives developers and DBAs the ability to test and debug
server-side programs using a graphical, dynamic environment. The types of programs
that can be debugged are SPL stored procedures, functions, triggers, and packages as well
as PL/pgSQL functions and triggers.

The Debugger is integrated with and invoked from the Postgres Enterprise Manager
client. There are two basic ways the Debugger can be used to test programs:

Standalone Debugging. The Debugger is used to start the program to be tested.

You supply any input parameter values required by the program and you can
immediately observe and step through the code of the program. Standalone
debugging is the typical method used for new programs and for initial problem
investigation.
In-Context Debugging. The program to be tested is initiated by an application
other than the Debugger. You first set a global breakpoint on the program to be
tested. The application that makes the first call to the program encounters the
global breakpoint. The application suspends execution at which point the
Debugger takes control of the called program. You can then observe and step
through the code of the called program as it runs within the context of the calling
application. After you have completely stepped through the code of the called
program in the Debugger, the suspended application resumes execution. In-
context debugging is useful if it is difficult to reproduce a problem using
standalone debugging due to complex interaction with the calling application.

The debugging tools and operations are the same whether using standalone or in-context
debugging. The difference is in how the program to be debugged is invoked.

The following sections discuss the features and functionality of the Debugger using the
standalone debugging method. The directions for starting the Debugger for in-context
debugging are discussed in Section 7.1.5.3.

7.1.1 Configuring the Debugger

Before using the Debugger, edit the postgresql.conf file (located in the data
subdirectory of your Advanced Server home directory), adding
$libdir/plugin_debugger to the libraries listed in the
shared_preload_libraries configuration parameter:

shared_preload_libraries = '$libdir/dbms_pipe,$libdir/edb_gen,$libdir/plugin_debugger'

After modifying the shared_preload_libraries parameter, you must restart the

database server.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 233

 EDB Postgres Advanced Server Guide

7.1.2 Starting the Debugger

You can use the Postgres Enterprise Manager (PEM) client to access the Debugger for
standalone debugging. To open the Debugger, highlight the name of the stored procedure
or function you wish to debug in the PEM Object browser panel. Then, navigate
through the Tools menu to the Debugging menu and select Debug from the submenu
as shown in Figure 7.1.

Figure 7.1 - Starting the Debugger from the Tools menu

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 234

 EDB Postgres Advanced Server Guide

You can also right-click on the name of the stored procedure or function in the PEM
client Object Browser, and select Debugging, and the Debug from the context menu
as shown in Figure 7.2.

Figure 7.2 - Starting the Debugger from the objects context menu

Note that triggers cannot be debugged using standalone debugging. Triggers must be
debugged using in-context debugging. See Section 7.1.5.3 for information on setting a
global breakpoint for in-context debugging.

To debug a package, highlight the specific procedure or function under the package node
of the package you wish to debug and follow the same directions as for stored procedures
and functions.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 235

 EDB Postgres Advanced Server Guide

7.1.3 The View Data Options Window

You can use the View Data Options window to pass parameter values when you are
standalone-debugging a program that expects parameters. When you start the debugger,
the View Data Options window opens automatically to display any IN or IN OUT
parameters expected by the program. If the program declares no IN or IN OUT
parameters, the View Data Options window does not open.

Figure 7.3 - The View Data Options window

Use the fields on the View Data Options window (shown in Figure 7.3) to provide a
value for each parameter:

The Name field contains the formal parameter name.

The Type field contains the parameter data type.

Check the Null? checkbox to indicate that the parameter is a NULL value.

Check the Expression checkbox if the Value field contains an expression.

The Value field contains the parameter value that will be passed to the program.

Check the Use default? checkbox to indicate that the program should use the
value in the Default Value field.

The Default Value field contains the default value of the parameter.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 236

 EDB Postgres Advanced Server Guide

Press the Enter key to select the next parameter in the list for data entry, or click on a
Value field to select the parameter for data entry.

If you are debugging a procedure or function that is a member of a package that has an
initialization section, check the Debug Package Initializer check box to instruct the
Debugger to step into the package initialization section, allowing you to debug the
initialization section code before debugging the procedure or function. If you do not
select the check box, the Debugger executes the package initialization section without
allowing you to see or step through the individual lines of code as they are executed.

After entering the desired parameter values, click the OK button to start the debugging
process. Click the Cancel button to terminate the Debugger and return control to the
PEM client.

Note: The View Data Options window does not open during in-context debugging.
Instead, the application calling the program to be debugged must supply any required
input parameter values.

When you have completed a full debugging cycle by stepping through the program code,
the View Data Options window re-opens, allowing you to enter new parameter values
and repeat the debugging cycle, or end the debugging session.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 237

 EDB Postgres Advanced Server Guide

7.1.4 Main Debugger Window

The Main Debugger window (see Figure 7.4) contains three panes:

the Program Body pane

the Stack pane
the Output pane

You can use the debugger menu bar or tool bar icons (located at the top of the debugger
window) to access debugging functions.

Figure 7.4 - The Main Debugger window

Status and error information is displayed in the status bar at the bottom of the Debugger
window.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 238

 EDB Postgres Advanced Server Guide

7.1.4.1 The Program Body Pane

The Program Body pane in the upper-left corner of the Debugger window displays the
source code of the program that is being debugged.

Figure 7.5 - The Program Body

Figure 7.5 shows that the Debugger is about to execute the SELECT statement. The green
indicator in the program body highlights the next statement to execute.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 239

 EDB Postgres Advanced Server Guide

7.1.4.2 The Stack Pane

The Stack pane displays a list of programs that are currently on the call stack (programs
that have been invoked but which have not yet completed). When a program is called,
the name of the program is added to the top of the list displayed in the Stack pane;
when the program ends, its name is removed from the list.

The Stack pane also displays information about program calls. The information includes:

The location of the call within the program

The call arguments
The name of the program being called

Reviewing the call stack can help you trace the course of execution through a series of
nested programs.

Figure 7.6 A debugged program calling a subprogram

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 240

 EDB Postgres Advanced Server Guide

Figure 7.6 shows that emp_query_caller is about to call a subprogram named

emp_query. emp_query_caller is currently at the top of the call stack.

After the call to emp_query executes, emp_query is displayed at the top of the Stack
pane, and its code is displayed in the Program Body frame (see Figure 7.7).

Figure 7.7 - Debugging the called subprogram

Upon completion of execution of the subprogram, control returns to the calling program
(public.emp_query_caller), now displayed at the top of the Stack pane in Figure
7.8.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 241

 EDB Postgres Advanced Server Guide

Figure 7.8 Control returns from debugged subprogram

Highlight an entry in the call stack to review detailed information about the selected entry
on the tabs in the Output pane. Using the call stack to navigate to another entry in the
call stack will not alter the line that is currently executing.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 242

 EDB Postgres Advanced Server Guide

7.1.4.3 The Output Pane

You can use tabs in the Output pane (see Figure 7.9) to view or modify parameter
values or local variables, or to view messages generated by RAISE INFO and function
results.

Figure 7.9 The DBMS Messages tab of the Output pane.

Each tab contains a different type of information:

The Parameters tab displays the current parameter values.

The Local Variables tab displays the value of any variables declared within
the program.

The DBMS Messages tab displays any results returned by the program as it
executes.

The Results tab displays program results (if applicable).

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 243

 EDB Postgres Advanced Server Guide

7.1.4.4 The Status Bar

The status bar (see Figure 7.10) displays a message when the Debugger pauses, when a
runtime error message is encountered, or when execution completes.

Figure 7.10 - The Status Bar, indicating Execution completed.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 244

 EDB Postgres Advanced Server Guide

7.1.5 Debugging a Program

You can perform the following operations to debug a program:

Step through the program one line at a time

Execute the program until you reach a breakpoint
View and change local variable values within the program

7.1.5.1 Stepping Through the Code

Use the tool bar icons to step through a program with the Debugger:

Use the Step Into icon to execute the line of code currently highlighted
by the green bar in the Program Body pane, and then pause execution. If the
executed code line is a call to a subprogram, the called subprogram is brought into the
Program Body pane, and the first executable line of code of the subprogram is
highlighted as the Debugger waits for you to perform an operation on the
subprogram.

Use the Step Over icon to execute a line of code, stepping over
any subprograms invoked by that line of code. The subprogram is executed, but not
debugged. If the subprogram contains a breakpoint, the debugger will stop at that
breakpoint.

Use the Continue icon to execute the line of code highlighted by the
green bar, and continue execution until either a breakpoint is encountered or the last
line of the program has been executed.

Figure 7.11 shows the locations of the Step Into, Step Over, and Continue icons on
the tool bar:

Figure 7.11 - The Step Into, Step Over, and Continue icons

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 245

 EDB Postgres Advanced Server Guide

The debugging operations are also accessible through the Debug menu, as shown in
Figure 7.12.

Figure 7.12 - Debug menu options

7.1.5.2 Using Breakpoints

As the Debugger executes a program, it pauses whenever it reaches a breakpoint. When
the Debugger pauses, you can observe or change local variables, or navigate to an entry
in the call stack to observe variables or set other breakpoints. The next step into, step
over, or continue operation forces the debugger to resume execution with the next line of
code following the breakpoint. There are two types of breakpoints:

Local Breakpoint - A local breakpoint can be set at any executable line of code within a
program. The Debugger pauses execution when it reaches a line where a local breakpoint
has been set.

Global Breakpoint - A global breakpoint will trigger when any session reaches that
breakpoint. Set a global breakpoint if you want to perform in-context debugging of a
program. When a global breakpoint is set on a program, the debugging session that set
the global breakpoint waits until that program is invoked in another session. A global
breakpoint can only be set by a superuser.

To create a local breakpoint, left-click in the grey shaded margin to the left of the line of
code where you want the local breakpoint set. The Debugger displays a red dot in the
margin, indicating a breakpoint has been set at the selected line of code (see Figure 7.13).

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 246

 EDB Postgres Advanced Server Guide

Figure 7.13 - Set a breakpoint by clicking in left-hand margin

You can also set a breakpoint by left-clicking in the Program Body to place your
cursor, and selecting Toggle Breakpoint from Debug menu or by clicking the
Toggle Breakpoint icon (see Figure 7.14). A red dot appears in the left-hand margin
indicating a breakpoint has been set as the line of code.

Figure 7.14 - The breakpoint control icons

You can set as many local breakpoints as desired. Local breakpoints remain in effect for
the duration of a debugging session until they are removed.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 247

 EDB Postgres Advanced Server Guide

Removing a Local Breakpoint

To remove a local breakpoint, you can:

Left click the mouse on the red breakpoint indicator in the left margin of the
Program Body pane. The red dot disappears, indicating that the breakpoint has
been removed.
Use your mouse to select the location of the breakpoint in the code body, and
select Toggle Breakpoint from Debug menu, or click the Toggle
Breakpoint icon.

You can remove all of the breakpoints from the program that currently appears in the
Program Body frame by selecting Clear all breakpoints from the Debug menu
(see Figure 7.15) or by clicking the Clear All Breakpoints icon.

Figure 7.15 - The breakpoint menu options

Note: When you perform any of the preceding actions, only the breakpoints in the
program that currently appears in the Program Body frame are removed. Breakpoints in
called subprograms or breakpoints in programs that call the program currently appearing
in the Program Body frame are not removed.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 248

 EDB Postgres Advanced Server Guide

7.1.5.3 Setting a Global Breakpoint for In-Context Debugging

To set a global breakpoint for in-context debugging, highlight the stored procedure,
function, or trigger on which you wish to set the breakpoint in the Object browser
panel. Navigate through the Tools menu to select Debugging, and then Set
Breakpoint (see Figure 7.16)

Figure 7.16 - Setting a global breakpoint from the Tools menu

Alternatively, you can right-click on the name of the stored procedure, function, or
trigger on which you wish to set a global breakpoint and select Debugging, then Set
Breakpoint from the context menu as shown in Figure 7.17.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 249

 EDB Postgres Advanced Server Guide

Figure 7.17 - Setting a global breakpoint from the object's context menu

To set a global breakpoint on a trigger, expand the table node that contains the trigger,
highlight the specific trigger you wish to debug, and follow the same directions as for
stored procedures and functions.

To set a global breakpoint in a package, highlight the specific procedure or function

under the package node of the package you wish to debug and follow the same directions
as for stored procedures and functions.

After you choose Set Breakpoint, the Debugger window opens and waits for an
application to call the program to be debugged (see Figure 7.18).

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 250

 EDB Postgres Advanced Server Guide

Figure 7.18 - Waiting for invocation of program to be debugged

In Figure 7.19, the EDB-PSQL client invokes the select_emp function (on which a
global breakpoint has been set).

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 251

 EDB Postgres Advanced Server Guide

Figure 7.19 - Application invoking program with a global breakpoint

The select_emp function does not complete until you step through the program in the
Debugger, which now appears as shown in Figure 7.20.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 252

 EDB Postgres Advanced Server Guide

Figure 7.20 - Program on which a global breakpoint has been set

You can now debug the program using any of the previously discussed operations such as
step into, step over, and continue, or set local breakpoints. When you have stepped
through execution of the program, the calling application (EDB-PSQL) regains control as
shown in Figure 7.21.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 253

 EDB Postgres Advanced Server Guide

Figure 7.21 - Application after debugging

The select_emp function completes execution and its output is displayed.

At this point, you can end the Debugger session by choosing Exit from the File menu.
If you do not end the Debugger session, the next application that invokes the program
will encounter the global breakpoint and the debugging cycle will begin again.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 254

 EDB Postgres Advanced Server Guide

7.1.5.4 Exiting the Debugger

To end a Debugger session and exit the Debugger, select Exit from File menu or press
Alt-F4 as shown by the following:

Figure 7.22 - Exiting from the Debugger

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 255

 EDB Postgres Advanced Server Guide

8 Performance Analysis and Tuning

Advanced Server provides various tools for performance analysis and tuning. These
features are described in this section.

8.1 Dynatune
Advanced Server supports dynamic tuning of the database server to make the optimal
usage of the system resources available on the host machine on which it is installed. The
two parameters that control this functionality are located in the postgresql.conf file.
These parameters are:

edb_dynatune
edb_dynatune_profile

8.1.1 edb_dynatune

edb_dynatune determines how much of the host system's resources are to be used by
the database server based upon the host machine's total available resources and the
intended usage of the host machine.

When Advanced Server is initially installed, the edb_dynatune parameter is set in

accordance with the selected usage of the host machine on which it was installed - i.e.,
development machine, mixed use machine, or dedicated server. For most purposes, there
is no need for the database administrator to adjust the various configuration parameters in
the postgresql.conf file in order to improve performance.

You can change the value of the edb_dynatune parameter after the initial installation of
Advanced Server by editing the postgresql.conf file. The postmaster must be
restarted in order for the new configuration to take effect.

The edb_dynatune parameter can be set to any integer value between 0 and 100,
inclusive. A value of 0, turns off the dynamic tuning feature thereby leaving the database
server resource usage totally under the control of the other configuration parameters in
the postgresql.conf file.

A low non-zero, value (e.g., 1 - 33) dedicates the least amount of the host machine's
resources to the database server. This setting would be used for a development machine
where many other applications are being used.

A value in the range of 34 - 66 dedicates a moderate amount of resources to the database

server. This setting might be used for a dedicated application server that may have a fixed
number of other applications running on the same machine as Advanced Server.
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 256
 EDB Postgres Advanced Server Guide

The highest values (e.g., 67 - 100) dedicate most of the server's resources to the database
server. This setting would be used for a host machine that is totally dedicated to running
Advanced Server.

Once a value of edb_dynatune is selected, database server performance can be further

fine-tuned by adjusting the other configuration parameters in the postgresql.conf
file. Any adjusted setting overrides the corresponding value chosen by edb_dynatune.
You can change the value of a parameter by un-commenting the configuration parameter,
specifying the desired value, and restarting the database server.

8.1.2 edb_dynatune_profile

The edb_dynatune_profile parameter is used to control tuning aspects based upon

the expected workload profile on the database server. This parameter takes effect upon
startup of the database server.

The possible values for edb_dynatune_profile are:

Value Usage
Recommended when the database server is processing heavy online transaction
oltp
processing workloads.
reporting Recommended for database servers used for heavy data reporting.
Recommended for servers that provide a mix of transaction processing and data
mixed
reporting.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 257

 EDB Postgres Advanced Server Guide

8.2 Infinite Cache

Note: Infinite Cache has been deprecated and may be removed in a future release. Please
contact your EnterpriseDB Account Manager or mailto:sales@enterprisedb.com for more
information.

Database performance is typically governed by two competing factors:

Memory access is fast; disk access is slow.

Memory space is scarce; disk space is abundant.

Advanced Server tries very hard to minimize disk I/O by keeping frequently used data in
memory. When the first server process starts, it creates an in-memory data structure
known as the buffer cache. The buffer cache is organized as a collection of 8K (8192
byte) pages: each page in the buffer cache corresponds to a page in some table or index.
The buffer cache is shared between all processes servicing a given database.

When you select a row from a table, Advanced Server reads the page that contains the
row into the shared buffer cache. If there isn't enough free space in the cache, Advanced
Server evicts some other page from the cache. If Advanced Server evicts a page that has
been modified, that data is written back out to disk; otherwise, it is simply discarded.
Index pages are cached in the shared buffer cache as well.

Figure 7.1 demonstrates the flow of data in a typical Advanced Server session:

Figure 7.1 Data Flow

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 258

 EDB Postgres Advanced Server Guide

A client application sends a query to the Postgres server and the server searches the
shared buffer cache for the required data. If the requested data is found in the cache, the
server immediately sends the data back to the client. If not, the server reads the page that
holds the data into the shared buffer cache, evicting one or more pages if necessary. If
the server decides to evict a page that has been modified, that page is written to disk.

As you can see, a query will execute much faster if the required data is found in the
shared buffer cache.

One way to improve performance is to increase the amount of memory that you can
devote to the shared buffer cache. However, most computers impose a strict limit on the
amount of RAM that you can install. To help circumvent this limit, Infinite Cache lets
you utilize memory from other computers connected to your network.

With Infinite Cache properly configured, Advanced Server will dedicate a portion of the
memory installed on each cache server as a secondary memory cache. When a client
application sends a query to the server, the server first searches the shared buffer cache
for the required data; if the requested data is not found in the cache, the server searches
for the necessary page in one of the cache servers.

Figure 7.2 shows the flow of data in an Advanced Server session with Infinite Cache:

Figure 7.2 Data flow with Infinite Cache

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 259

 EDB Postgres Advanced Server Guide

When a client application sends a query to the server, the server searches the shared
buffer cache for the required data. If the requested data is found in the cache, the server
immediately sends the data back to the client. If not, the server sends a request for the
page to a specific cache server; if the cache server holds a copy of the page it sends the
data back to the server and the server copies the page into the shared buffer cache. If the
required page is not found in the primary cache (the shared buffer cache) or in the
secondary cache (the cloud of cache servers), Advanced Server must read the page from
disk. Infinite Cache improves performance by utilizing RAM from other computers on
your network in order to avoid reading frequently accessed data from disk.

Updating the Cache Node Configuration

You can add or remove cache servers without restarting the database server by adding or
deleting cache nodes from the list defined in the edb_icache_servers configuration
parameter. For more information about changing the configuration parameter, see
Section 8.2.2.2.

When you add one or more cache nodes, the server re-allocates the cache, dividing the
cache evenly amongst the servers; each of the existing cache servers loses a percentage of
the information that they have cached. You can calculate the percentage of the cache that
remains valid with the following formula:

(existing_nodes * 100) / (existing_nodes + new_nodes)

For example, if an Advanced Server installation with three existing cache nodes adds an
additional cache node, 75% of the existing cache remains valid after the reconfiguration.

If cache nodes are removed from a server, the data that has been stored on the remaining
cache nodes is preserved. If one cache server is removed from a set of five cache servers,
Advanced Server preserves the 80% of the distributed cache that is stored on the four
remaining cache nodes.

When you change the cache server configuration (by adding or removing cache servers),
the portion of the cache configuration that is preserved is not re-written unless the cache
is completely re-warmed using the edb_icache_warm() function or
edb_icache_warm utility. If you do not re-warm the cache servers, new cache servers
will accrue cache data as queries are performed on the server.

Infinite Cache Offers a Second Performance Advantage: Compression.

Without Infinite Cache, Advanced Server will read each page from disk as an 8K chunk;
when a page resides in the shared buffer cache, it consumes 8K of RAM. With Infinite
Cache, Postgres can compress each page before sending it to a cache server. A
compressed page can take significantly less room in the secondary cache, making more
space available for other data and effectively increasing the size of the cache. A

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 260

 EDB Postgres Advanced Server Guide

compressed page consumes less network bandwidth as well, decreasing the amount of
time required to retrieve a page from the secondary cache.

The fact that Infinite Cache can compress each page may make it attractive to configure a
secondary cache server on the same computer that runs your Postgres server. If, for
example, your computer is configured with 6GB of RAM, you may want to allocate a
smaller amount (say 1GB) for the primary cache (the shared buffer cache) and a larger
amount (4GB) to the secondary cache (Infinite Cache), reserving 1GB for the operating
system. Since the secondary cache resides on the same computer, there is very little
overhead involved in moving data between the primary and secondary cache. All data
stored in the Infinite Cache is compressed so the secondary cache can hold many more
pages than would fit into the (uncompressed) shared buffer cache. If you had allocated
5GB to the shared buffer cache, the cache could hold no more than 65000 pages
(approximately). By assigning 4GB of memory to Infinite Cache, the cache may be able
to hold 130000 pages (at 2x compression), 195000 pages (at 3x compression) or more.
The compression factor that you achieve is determined by the amount of redundancy in
the data itself and the edb_icache_compression_level parameter.

To use Infinite Cache, you must specify a list of one or more cache servers (computers on
your network) and start the edb_icache daemon on each of those servers.

Infinite Cache is supported on Linux, HPUX and Solaris systems only.

Please Note: Infinite Cache and the effective_io_concurrency parameter can

potentially interfere with each other. You should disable asynchronous I/O requests (by
setting the value of effective_io_concurrency to 0 in the postgresql.conf file)
if you enable the Infinite Cache feature.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 261

 EDB Postgres Advanced Server Guide

8.2.1 Installing Infinite Cache

Advanced Server includes Infinite Cache functionality as part of a standard installation

with either the graphical installer or the RPM installer. You can also optionally install
only the Infinite Cache daemon on a supporting cache server.

For information about using the RPM packages to install Infinite Cache, please see the
EDB Postgres Advanced Server Installation Guide available at:

http://www.enterprisedb.com/products-services-
training/products/documentation/enterpriseedition

To use the graphical installer to install Advanced Server with Infinite Cache
functionality, confirm that the box next to the Database Server option (located on the
Select Components dialog, shown in Figure 7.3) is selected when running the
installation wizard.

Figure 7.3: The Select Components dialog.

The Database Server option installs the following Infinite Cache components:

The edb-icache service script.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 262

 EDB Postgres Advanced Server Guide

The Infinite Cache configuration file (edb-icache).

A command line tool that allows you to pre-load the cache servers (edb-
icache-warm).
The edb_icache libraries (code libraries required by the edb-icache
daemon).

The graphical installation wizard can selectively install only the Infinite Cache daemon
on a cache server. To install the edb-icache daemon on a cache server, deploy the
installation wizard on the machine hosting the cache; when the Setup: Select
Components window opens, de-select all options except Infinite Cache (as shown
in Figure 7.4).

Figure 8.4: Installing only the Infinite Cache Daemon.

The Infinite Cache Daemon option installs the following:

The edb-icache service script.

The Infinite Cache configuration file (edb-icache).
A command line tool that allows you to pre-load the cache servers (edb-
icache-warm).
The edb_icache libraries (code libraries required by the edb-icache
daemon).

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 263

 EDB Postgres Advanced Server Guide

8.2.2 Configuring the Infinite Cache Server

Configuring Infinite Cache is a three-step process:

Specify Infinite Cache server settings in the Infinite Cache configuration file.
Modify the Advanced Server postgresql.conf file, enabling Infinite Cache,
and specifying connection and compression settings.
Start the Infinite Cache service.

8.2.2.1 Modifying Infinite Cache Settings

The Infinite Cache configuration file is named edb-icache, and contains two
parameters and their associated values:

PORT=11211
CACHESIZE=500

To modify a parameter, open the edb-icache file (located in the

/opt/edb/icache/etc directory) with your editor of choice, and modify the
parameter values:

PORT

Use the PORT variable to specify the port where Infinite Cache will listen for
connections from Advanced Server.

CACHESIZE

Use the CACHESIZE variable to specify the size of the cache (in MB).

8.2.2.2 Enabling Infinite Cache

The postgresql.conf file includes three configuration parameters that control the
behavior of Infinite Cache. The postgresql.conf file is read each time you start the
Advanced Server database server. To modify a parameter, open the postgresql.conf
file (located in the $PGDATA directory) with your editor of choice, and edit the section of
the configuration file shown below:

# - Infinite Cache
#edb_enable_icache = off
#edb_icache_servers = '' #'host1:port1,host2,ip3:port3,ip4'
#edb_icache_compression_level = 6

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 264

 EDB Postgres Advanced Server Guide

Lines that begin with a pound sign (#) are treated as comments; to enable a given
parameter, remove the pound sign and specify a value for the parameter. When you've
updated and saved the configuration file, restart the database server for the changes to
take effect.

edb_enable_icache

Use the edb_enable_icache parameter to enable or disable Infinite Cache.

When edb_enable_icache is set to on, Infinite Cache is enabled; if the
parameter is set to off, Infinite Cache is disabled.

If you set edb_enable_icache to on, you must also specify a list of cache
servers by setting the edb_icache_servers parameter (described in the next
section).

The default value of edb_enable_icache is off.

edb_icache_servers

The edb_icache_servers parameter specifies a list of one or more servers

with active edb-icache daemons. edb_icache_servers is a string value that
takes the form of a comma-separated list of hostname:port pairs. You can specify
each pair in any of the following forms:

hostname
IP-address
hostname:portnumber
IP-address:portnumber

If you do not specify a port number, Infinite Cache assumes that the cache server
is listening at port 11211. This configuration parameter will take effect only if
edb_enable_icache is set to on. Use the edb_icache_servers parameter
to specify a maximum of 128 cache nodes.

edb_icache_compression_level

The edb_icache_compression_level parameter controls the compression

level that is applied to each page before storing it in the distributed Infinite Cache.
This parameter must be an integer in the range 0 to 9.

A compression level of 0 disables compression; it uses no CPU time for

compression, but requires more storage space and network resources to
process.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 265

 EDB Postgres Advanced Server Guide

A compression level of 9 invokes the maximum amount of compression; it

increases the load on the CPU, but less data flows across the network, so
network demand is reduced. Each page takes less room on the Infinite
Cache server, so memory requirements are reduced.

A compression level of 5 or 6 is a reasonable compromise between the

amount of compression received and the amount of CPU time invested.

By default, edb_icache_compression_level is set to 6.

When Advanced Server reads data from disk, it typically reads the data in 8K
increments. If edb_icache_compression_level is set to 0, each time
Advanced Server sends an 8K page to the Infinite Cache server that page is stored
(uncompressed) in 8K of cache memory. If the
edb_icache_compression_level parameter is set to 9, Advanced Server
applies the maximum compression possible before sending it to the Infinite Cache
server, so a page that previously took 8K of cached memory might take 2K of
cached memory. Exact compression numbers are difficult to predict, as they are
dependent on the nature of the data on each page.

The compression level must be set by the superuser and can be changed for the
current session while the server is running. The following command disables the
compression mechanism for the currently active session:

SET edb_icache_compression_level = 0

The following example shows a typical collection of Infinite Cache settings:

edb_enable_icache = on
edb_icache_servers = 'localhost,192.168.2.1:11200,192.168.2.2'
edb_icache_compression_level = 6

Please Note: Infinite Cache and the effective_io_concurrency parameter can

potentially interfere with each other. You should disable asynchronous I/O requests (by
setting the value of effective_io_concurrency to 0 in the postgresql.conf file)
if you enable the Infinite Cache feature. By default, effective_io_concurrency is
set to 1.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 266

 EDB Postgres Advanced Server Guide

8.2.2.3 Controlling the Infinite Cache Server

Linux

On Linux, the Infinite Cache service script is named edb-icache. The service script
resides in the /etc/init.d directory. You can control the Infinite Cache service, or
check the status of the service with the following command:

/etc/init.d/edb-icache action

Where action specifies:

start to start the service.

stop to stop the service
restart to stop and then start the service.
status to return the status of the service.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 267

 EDB Postgres Advanced Server Guide

8.2.3 Dynamically Modifying Infinite Cache Server Nodes

You can dynamically modify the Infinite Cache server nodes; to change the Infinite
Cache server configuration, use the edb_icache_servers parameter in the
postgresql.conf file to:

specify additional cache information to add a server/s.

delete server information to remove a server/s.

specify additional server information and delete existing server information to

both add and delete servers during the same reload operation.

After updating the edb_icache_servers parameter in the postgresql.conf file, you

must reload the configuration parameters for the changes to take effect. You can use the
menu-driven reload options (navigate through the EDB Postgres menu to the
Advanced Server 10 menu; continue to the Expert Configuration menu, and
select the Reload Configuration option). If prompted, enter your password to reload
the configuration parameters.

Alternatively, you can use the pg_ctl reload command to update the server's
configuration parameters at the command line:

pg_ctl reload -D data_directory

Where data_directory specifies the complete path to the data directory.

Please Note: If the server detects a problem with the value specified for the
edb_icache_servers parameter during a server reload, it will ignore changes to the
parameter and use the last valid parameter value. If you are performing a server
restart, and the parameter contains an invalid value, the server will return an error.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 268

 EDB Postgres Advanced Server Guide

8.2.4 Controlling the edb-icache Daemons

edb-icache is a high-performance memory caching daemon that distributes and stores

data in shared buffers. The server transparently interacts with edb-icache daemon to
store and retrieve data.

Before starting the database server, the edb-icache daemon must be running on each
server node. Log into each server and start the edb-icache server (on that host) by
issuing the following command:

# edb-icache -u enterprisedb -d -m 1024

Where:

-u

-u specifies the user name

-m

-m specifies the amount of memory to be used by edb-icache. The default is

64MB.

-d

-d designates that the service should run in the background

To gracefully kill an edb-icache daemon (close any in-use files, flush buffers, and
exit), execute the command:

# killall -TERM edb-icache

If the edb-icache daemon refuses to die, you may need to use the following command:

# killall -KILL edb-icache

8.2.4.1 Command Line Options

To view the command line options for the edb-icache daemon, use the following
command:

# /opt/edb/icache/bin/edb-icache -h

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 269

 EDB Postgres Advanced Server Guide

The command line options are:

Parameter Description
-p <port_number> The TCP port number the Infinite Cache daemon is listening on. The default is
11211.
-U <UDP_number> The UDP port number the Infinite Cache daemon is listening on. The default is
0 (off).
-s <pathname> The Unix socket pathname the Infinite Cache daemon is listening on. If
included, the server limits access to the host on which the Infinite Cache
daemon is running, and disables network support for Infinite Cache.
-a <mask> The access mask for the Unix socket, in octal form. The default value is 0700.
-l <ip_addr> Specifies the IP address that the daemon is listening on. If an individual address
is not specified, the default value is INDRR_ANY; all IP addresses assigned to
the resource are available to the daemon.
-d Run as a daemon.
-r Maximize core file limit.
-u <username> Assume the identity of the specified user (when run as root).
-m <numeric> Max memory to use for items in megabytes. Default is 64 MB.
-M Return error on memory exhausted (rather than removing items).
-c <numeric> Max simultaneous connections. Default is 1024.
-k Lock down all paged memory. Note that there is a limit on how much memory
you may lock. Trying to allocate more than that would fail, so be sure you set
the limit correctly for the user you started the daemon with (not for -u
<username> user; under sh this is done with 'ulimit -S -l NUM_KB').
-v Verbose (print errors/warnings while in event loop).
-vv Very verbose (include client commands and responses).
-vvv Extremely verbose (also print internal state transitions).
-h Print the help text and exit.
-i Print memcached and libevent licenses.
-P <file> Save PID in <file>, only used with -d option.
-f <factor> Chunk size growth factor. Default value is 1.25.
-n <bytes> Minimum space allocated for key+value+flags. Default is 48.
-L Use large memory pages (if available). Increasing the memory page size could
reduce the number of transition look-aside buffer misses and improve the
performance. To get large pages from the OS, Infinite Cache will allocate the
total item-cache in one large chunk.
-D <char> Use <char> as the delimiter between key prefixes and IDs. This is used for per-
prefix stats reporting. The default is":" (colon).
If this option is specified, stats collection is enabled automatically; if not, then it
may be enabled by sending the stats detail on command to the server.
-t <num> Specifies the number of threads to use. Default is 4.
-R Maximum number of requests per event; this parameter limits the number of
requests process for a given connection to prevent starvation, default is 20.
-C Disable use of CAS (check and set).
-b Specifies the backlog queue limit, default is 1024.
-B Specifies the binding protocol. Possible values are ascii, binary or auto;
default value is auto.
-I Override the size of each slab page. Specifies the max item size; default 1 MB,
minimum size is 1 k, maximum is 128 MB).

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 270

 EDB Postgres Advanced Server Guide

8.2.4.2 edb-icache-tool

edb-icache-tool provides a command line interface that queries the edb-icache daemon
to retrieve statistical information about a specific cache node. The syntax is:

edb-icache-tool <host[:port]> stats

host specifies the address of the host that you are querying.
port specifies the port that the daemon is listening on.

edb-icache-tool retrieves the statistics described in the following table:

Statistic Description
accepting_conns Will this server accept new connection(s)? 1 if yes, otherwise 0.
auth_cmds Number of authentication commands handled by this server, success or
failure.
auth_errors Number of failed authentications.
bytes Total number of bytes in use.
bytes_read Total number of bytes received by this server (from the network).
bytes_written Total number of bytes sent by this server (to the network).
cas_badval Number of keys that have been compared and swapped by this server but
the comparison (original) value did not match the supplied value.
cas_hits Number of keys that have been compared and swapped by this server and
found present.
cas_misses Number of keys that have been compared and swapped by this server and
not found.
cmd_flush Cumulative number of flush requests sent to this server.
cmd_get Cumulative number of read requests sent to this server.
cmd_set Cumulative number of write requests sent to this server.
conn_yields Number of times any connection yielded to another due to hitting the edb-
icache -R limit.
connection_structures Number of connection structures allocated by the server.
curr_connections Number of open connections.
curr_items Number of items currently stored by the server.
decr_hits Number of decrement requests satisfied by this server.
decr_misses Number of decrement requests not satisfied by this server.
delete_hits Number of delete requests satisfied by this server.
delete_misses Number of delete requests not satisfied by this server.
evictions Number of valid items removed from cache to free memory for new items.
get_hits Number of read requests satisfied by this server.
get_misses Number of read requests not satisfied by this server.
incr_hits Number of increment requests satisfied by this server.
incr_misses Number of increment requests not satisfied by this server.
limit_maxbytes Number of bytes allocated on this server for storage.
listen_disabled_num Cumulative number of times this server has hit its connection limit.
pid Process ID (on cache server).
pointer_size Default pointer size on host OS (usually 32 or 64).
reclaimed Number of times an entry was stored using memory from an expired entry.
rusage_user Accumulated user time for this process (seconds.microseconds).
rusage_system Accumulated system time for this process (seconds.microseconds).
threads Number of worker threads requested.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 271

 EDB Postgres Advanced Server Guide

total_time Number of seconds since this server's base date (usually midnight, January
1, 1970, UTC).
total_connections Total number of connections opened since the server started running.
total_items Total number of items stored by this server (cumulative).
uptime Amount of time that server has been active.
version edb-icache version.

In the following example, edb-icache-tool retrieves statistical information about an

Infinite Cache server located at the address, 192.168.23.85 and listening on port
11213:

# edb-icache-tool 192.168.23.85:11213 stats

Field Value
accepting_conns 1
auth_cmds 0
auth_errors 0
bytes 52901223
bytes_read 188383848
bytes_written 60510385
cas_badval 0
cas_hits 0
cas_misses 0
cmd_flush 1
cmd_get 53139
cmd_set 229120
conn_yields 0
connection_structures 34
curr_connections 13
curr_items 54953
decr_hits 0
decr_misses 0
delete_hits 0
delete_misses 0
evictions 0
get_hits 52784
get_misses 355
incr_hits 0
incr_misses 0
limit_maxbytes 314572800
listen_disabled_num 0
pid 7226
pointer_size 32
reclaimed 0
rusage_system 10.676667
rusage_user 3.068191
threads 4
time 1320919080
total_connections 111
total_items 229120
uptime 7649
version 1.4.5

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 272

 EDB Postgres Advanced Server Guide

8.2.5 Warming the edb-icache Servers

When the server starts, the primary and secondary caches are empty. When Advanced
Server processes a client request, the server reads the required data from disk and stores a
copy in each cache. You can improve server performance by warming (or pre-loading)
the data into the memory cache before a client asks for it.

There are two advantages to warming the cache. Advanced Server will find data in the
cache the first time it is requested by a client application, instead of waiting for it to be
read from disk. Also, manually warming the cache with the data that your applications
are most likely to need saves time by avoiding future random disk reads. If you don't
warm the cache at startup, Advanced Server performance may not reach full speed until
the client applications happen to load commonly used data into the cache.

There are several ways to load pages to warm the Infinite Cache server nodes. You can:

Use the edb_icache_warm utility to warm the caches from the command line.

Use the edb_icache_warm() function from within edb-psql.

Use the edb_icache_warm() function via scripts to warm the cache.

While it is not necessary to re-warm the cache after making changes to an existing cache
configuration, re-warming the cache can improve performance by bringing the new
configuration of cache servers up-to-date.

8.2.5.1 The edb_icache_warm() Function

The edb_icache_warm() function comes in two variations; the first variation warms
not only the table, but any indexes associated with the table. If you use the second
variation, you must make additional calls to warm any associated indexes.

The first form of the edb_icache_warm() function warms the given table and any
associated indexes into the cache. The signature is:

edb_icache_warm(table_name)

You may specify table_name as a table name, OID, or regclass value.

# edb-psql edb -c "select edb_icache_warm('accounts')"

When you call the first form of edb_icache_warm(), Advanced Server reads each
page in the given table, compresses the page (if configured to do so), and then sends the

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 273

 EDB Postgres Advanced Server Guide

compressed data to an Infinite Cache server. edb_icache_warm() also reads,

compresses, and caches each page in each index defined for the given table.

The second form of the edb_icache_warm() function warms the pages that contain the
specified range of bytes into the cache. The signature of the second form is:

edb_icache_warm(table-spec, startbyte, endbyte):

You must make subsequent calls to specify indexes separately when using this form of
the edb_icache_warm() function.

# edb-psql edb -c "select edb_icache_warm('accounts', 1, 10000)"

The edb_icache_warm() function is typically called by a utility program (such as the

edb_icache_warm utility) to spread the warming process among multiple processes
that operate in parallel.

8.2.5.2 Using the edb_icache_warm Utility

You can use the edb_icache_warm command-line utility to load the cache servers with
specified tables, allowing fast access to relevant data from the cache.

The syntax for edb_icache_warm is:

# edb_icache_warm d database t tablename

The only required parameter is tablename. tablename can be specified with or

without the -t option. All other parameters are optional; if omitted, default values are
inferred from Advanced Server environment variables.

The options for edb_icache_warm are:

Option Variable Description

-h Hostname The name of the host running Advanced Server. Include this parameter if
you are running Advanced Server on a remote host.
The default value is PGHOST.
-p Portname Port in use by Advanced Server. Default value is PGPORT.
-j process count Number of (parallel) processes used to warm the cache. The default value
is 1.
-U Username The Advanced Server username. Unless specified, this defaults to
PGUSER.
-d Database The name of database containing the tables to be warmed. Default value is
PGDATABASE.
-t Tablename Name of table to be warmed. The index for the table is also warmed.
Required.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 274

 EDB Postgres Advanced Server Guide

8.2.6 Retrieving Statistics from Infinite Cache

8.2.6.1 Using edb_icache_stats()

You can view Infinite Cache statistics by using the edb_icache_stats() function at
the edb-psql command line (or any other query tool). The edb_icache_stats()
function returns a result set that reflects the state of an Infinite Cache node or nodes and
the related usage statistics. The result set includes:

Statistic Description
hostname Host name (or IP address) of server
Port Port number at which edb-icache daemon is listening
State Health of this server
write_failures Number of write failures
Bytes Total number of bytes in use
bytes_read Total number of bytes received by this server (from the network)
bytes_written Total number of bytes sent by this server (to the network)
cmd_get Cumulative number of read requests sent to this server
cmd_set Cumulative number of write requests sent to this server
connection_structures Number of connection structures allocated by the server
curr_connections Number of open connections
curr_items Number of items currently stored by the server
Evictions Number of valid items removed from cache to free memory for new items
get_hits Number of read requests satisfied by this server
get_misses Number of read requests not satisfied by this server
limit_maxbytes Number of bytes allocated on this server for storage
Pid Process ID (on cache server)
pointer_size Default pointer size on host OS (usually 32 or 64)
rusage_user Accumulated user time for this process (seconds.microseconds)
rusage_system Accumulated system time for this process (seconds.microseconds)
Threads Number of worker threads requested
total_time Number of seconds since this server's base date (usually midnight, January
1, 1970, UTC)
total_connections Total number of connections opened since the server started running
total_items Total number of items stored by this server (cumulative)
Uptime Amount of time that server has been active
Version edb-icache version

You can use SQL queries to view Infinite Cache statistics. To view the server status of
all Infinite Cache nodes:

SELECT hostname, port, state FROM edb_icache_stats()

hostname | port | state

---------------+-------+--------
192.168.23.85 | 11211 | UNHEALTHY
192.168.23.85 | 11212 | ACTIVE
(2 rows)

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 275

 EDB Postgres Advanced Server Guide

Use the following command to view complete statistics (shown here using edb-psql's
expanded display mode, \x) for a specified node:

SELECT * FROM edb_icache_stats() WHERE hostname = '192.168.23.85:11211'

-[RECORD 1]-----------+--------------
hostname | 192.168.23.85
port | 11211
state | ACTIVE
write_failures | 0
bytes | 225029460
bytes_read | 225728252
bytes_written | 192806774
cmd_get | 23313
cmd_set | 27088
connection_structures | 53
curr_connections | 3
curr_items | 27088
evictions | 0
get_hits | 23266
get_misses | 47
limit_maxbytes | 805306368
pid | 4240
pointer_size | 32
rusage_user | 0.481926
rusage_system | 1.583759
threads | 1
total_time | 1242199782
total_connections | 66
total_items | 27088
uptime | 714
version | 1.2.6

8.2.6.2 edb_icache_server_list

The edb_icache_server_list view exposes information about the status and health
of all Infinite Cache servers listed in the edb_icache_servers GUC. The
edb_icache_server_list view is created using the edb_icache stats() API.
The view exposes the following information for each server:

Statistic Description
Hostname Host name (or IP address) of server
Port Port number at which edb-icache daemon is listening
State Health of this server
write_failures Number of write failures
total_memory Number of bytes allocated to the cache on this server
memory_used Number of bytes currently used by the cache
memory_free Number of unused bytes remaining in the cache
hit_ratio Percentage of cache hits

The state column will contain one of the following four values, reflecting the health of
the given server:

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 276

 EDB Postgres Advanced Server Guide

Server State Description

Active The server is known to be up and running.
Unhealthy An error occurred while interacting with the cache server. Postgres will
attempt to re-establish the connection with the server.
Offline Postgres can no longer contact the given server.
Manual Offline You have taken the server offline with the edb_icache_server_enable()
function.

Use the following SELECT statement to return the health of each node in the Infinite
Cache server farm:

SELECT hostname, port, state FROM edb_icache_server_list

hostname | port | state

---------------+-------+-------
192.168.23.85 | 11211 | ACTIVE
192.168.23.85 | 11212 | ACTIVE
(2 rows)

Use the following command to view complete details about a specific Infinite Cache
node (shown here using edb-psql's \x expanded-view option):

SELECT * FROM edb_icache_server_list WHERE hostname = '192.168.23.85:11211'

-[RECORD 1]-----------+--------------
hostname | 192.168.23.85
port | 11211
state | ACTIVE
write_failures | 0
total_memory | 805306368
memory_used | 225029460
memory_free | 580276908
hit_ratio | 99.79

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 277

 EDB Postgres Advanced Server Guide

8.2.7 Retrieving Table Statistics

Advanced Server provides six system views that contain statistical information on a per-
table basis. The views are:

pg_statio_all_tables
pg_statio_sys_tables
pg_statio_user_tables
pg_statio_all_indexes
pg_statio_sys_indexes
pg_statio_user_indexes

You can use standard SQL queries to view and compare the information stored in the
views. The views contain information that will allow you to observe the effectiveness of
the Advanced Server buffer cache and the icache servers.

8.2.7.1 pg_statio_all_tables

The pg_statio_all_tables view contains one row for each table in the database.
The view contains the following information:

Column Name Description

relid The OID of the table.
schemaname The name of the schema that the table resides in.
relname The name of the table.
heap_blks_read The number of heap blocks read.
heap_blks_hit The number of heap blocks hit.
heap_blks_icache_hit The number of heap blocks found on an icache server.
idx_blks_read The number of index blocks read.
idx_blks_hit The number of index blocks hit.
idx_blks_icache_hit The number of index blocks found on an icache server.
toast_blks_read The number of toast blocks read.
toast_blks_hit The number of toast blocks hit.
toast_blks_icache_hit The number of toast blocks found on an icache server.
tidx_blks_read The number of index toast blocks read.
tidx_blks_hit The number of index toast blocks hit.
tidx_blks_icache_hit The number of index toast blocks found on an icache server.

You can execute a simple query to view performance statistics for a specific table:

SELECT * FROM pg_statio_all_tables WHERE relname=jobhist;

-[ RECORD 1 ]---------+---------
relid | 16402
schemaname | public
relname | jobhist
heap_blks_read | 1
heap_blks_hit | 51
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 278
 EDB Postgres Advanced Server Guide

heap_blks_icache_hit | 0
idx_blks_read | 2
idx_blks_hit | 17
idx_blks_icache_hit | 0
toast_blks_read |
toast_blks_hit |
toast_blks_icache_hit |
tidx_blks_read |
tidx_blks_hit |
tidx_blks_icache_hit |

Or, you can view the statistics by activity level. The following example displays the
statistics for the ten tables that have the greatest heap_blks_icache_hit activity:

SELECT * FROM pg_statio_all_tables ORDER BY heap_blks_icache_hit DESC LIMIT

10;

relid schemaname relname

heap_blks_read heap_blks_hit heap_blks_icache_hit
idx_blks_read idx_blks_hit idx_blks_icache_hit
toast_blks_read toast_blks_hit toast_blks_icache_hit
tidx_blks_read tidx_blks_hit tidx_blks_icache_hit
-----------------------------------------------------------------------------
16390 public pgbench_accounts
264105 71150 81498
13171 282541 18053

1259 pg_catalog pg_class

22 2904 18
14 3449 11

1249 pg_catalog pg_attribute

49 1619 16
17 2841 13

1255 pg_catalog pg_proc

38 276 11
33 682 16
0 0 0
0 0 0

2619 pg_catalog pg_statistic

20 295 8
4 436 4
0 0 0
0 0 0

2617 pg_catalog pg_operator

20 293 8
19 791 10

2602 pg_catalog pg_amop

10 721 6
13 1154 13

2610 pg_catalog pg_index

10 633 6
8 719 8

1247 pg_catalog pg_type

17 235 5
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 279
 EDB Postgres Advanced Server Guide

12 433 4

2615 pg_catalog pg_namespace

4 260 4
6 330 4
0 0 0
0 0 0
(10 rows)

8.2.7.2 pg_statio_sys_tables

The pg_statio_sys_tables view contains one row for each table in a system-defined
schema. The statistical information included in this view is the same as for
pg_statio_all_tables.

8.2.7.3 pg_statio_user_tables

The pg_statio_user_tables view contains one row for each table in a user-defined
schema. The statistical information in this view is the same as for
pg_statio_all_tables.

8.2.7.4 pg_statio_all_indexes

The pg_statio_all_indexes view contains one row for each index in the current
database. The view contains the following information:

Column Name Description

relid The OID of the indexed table
indexrelid The OID of the index.
schemaname The name of the schema that the table resides in.
relname The name of the table.
indexrelname The name of the index
idx_blks_read The number of index blocks read.
idx_blks_hit The number of index blocks hit.
idx_blks_icache_hit The number of index blocks found on an icache server.

You can execute a simple query to view performance statistics for the indexes on a
specific table:

SELECT * FROM pg_statio_all_indexes WHERE relname=pg_attribute;

-[ RECORD 1 ]---------+---------
relid | 1249
indexrelid | 2658
schemaname | pg_catalog
relname | pg_attribute
indexrelname | pg_attribute_relid_attnam_index
idx_blks_read | 10
idx_blks_hit | 1200
idx_blks_icache_hit | 0
-[ RECORD 2 ]---------+---------
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 280
 EDB Postgres Advanced Server Guide

relid | 1249
indexrelid | 2659
schemaname | pg_catalog
relname | pg_attribute
indexrelname | pg_attribute_relid_attnum_index
idx_blks_read | 12
idx_blks_hit | 3917
idx_blks_icache_hit | 0

The result set from the query includes the statistical information for two indexes; the
pg_attribute table has two indexes.

You can also view the statistics by activity level. The following example displays the
statistics for the ten indexes that have the greatest idx_blks_icache_hit activity:

SELECT * FROM pg_statio_all_indexes ORDER BY idx_blks_icache_hit DESC LIMIT

10;

relid indexrelid schemaname relname

indexrelname idx_blks_read idx_blks_hit idx_blks_icache_hit
-----------------------------------------------------------------------------
16390 16401 public pgbench_accounts
pgbench_accounts_pkey 13171 282541 18053

1249 2659 pg_catalog pg_attribute

pg_attr_relid_attnum_index 14 2749 13

1255 2690 pg_catalog proc

pg_proc_oid_index 16 580 12

1259 2663 pg_catalog pg_class

pg_class_relname_nsp_index 10 2019 7

2602 2654 pg_catalog pg_amop

pg_amop_opr_fam_index 7 453 7

2603 2655 pg_catalog pg_amproc

pg_amproc_fam_proc_index 6 605 6

2617 2688 pg_catalog pg_operator

pg_operator_oid_index 7 452 6

2602 2653 pg_catalog pg_amop

pg_amop_fam_strat_index 6 701 6

2615 2684 pg_catalog pg_namespace

pg_namespace_nspname_index 4 328 4

1262 2672 pg_catalog pg_database

pg_database_oid_index 4 254 4

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 281

 EDB Postgres Advanced Server Guide

8.2.7.5 pg_statio_sys_indexes

The pg_statio_sys_indexes view contains one row for each index on the system
tables. The statistical information in this view is the same as in
pg_statio_all_indexes.

8.2.7.6 pg_statio_user_indexes

The pg_statio_user_indexes view contains one row for each index on a table that
resides in a user-defined schema. The statistical information in this view is the same as
in pg_statio_all_indexes.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 282

 EDB Postgres Advanced Server Guide

8.2.8 edb_icache_server_enable()

You can use the edb_icache_server_enable() function to take the Infinite Cache
server offline for maintenance or other planned downtime. The syntax is:

void edb_icache_server_enable(host TEXT, port INTEGER, online BOOL)

host specifies the host that you want to disable. The host name may be specified by
name or numeric address.

port specifies the port number that the Infinite Cache server is listening on.

online specifies the state of the Infinite Cache server. The value of online must be true
or false.

To take a server offline, specify the host that you want to disable, the port number that
the Infinite Cache server is listening on, and false. To bring the Infinite Cache server
back online, specify the host name and port number, and pass a value of true.

The state of a server taken offline with the edb_icache_server_enable() function is

MANUAL OFFLINE. Advanced Server will not automatically reconnect to an Infinite
Cache server that you have taken offline with edb_icache_server_enable(...,
false); you must bring the server back online by calling
edb_icache_server_enable(..., true).

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 283

 EDB Postgres Advanced Server Guide

8.2.9 Infinite Cache Log Entries

When you start Advanced Server, a message that includes Infinite Cache status, cache
node count and cache node size is written to the server log. The following example
shows the server log for an active Infinite Cache installation with two 750 MB cache
servers:

** EnterpriseDB Dynamic Tuning Agent**************************************

* System Utilization: 66 % *
* Autovacuum Naptime: 60 Seconds *
* Infinite Cache: on *
* Infinite Cache Servers: 2 *
* Infinite Cache Size: 1.500 GB *
**************************************************************************

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 284

 EDB Postgres Advanced Server Guide

8.2.10 Allocating Memory to the Cache Servers

As mentioned earlier in this document, each computer imposes a limit on the amount of
physical memory that you can install. However, modern operating systems typically
simulate a larger address space so that programs can transparently access more memory
than is actually installed. This "virtual memory" allows a computer to run multiple
programs that may simultaneously require more memory than is physically available.
For example, you may run an e-mail client, a web browser, and a database server which
each require 1GB of memory on a machine that contains only 2GB of physical RAM.
When the operating system runs out of physical memory, it starts swapping bits and
pieces of the currently running programs to disk to make room to satisfy your current
demand for memory.

This can bring your system to a grinding halt.

Since the primary goal of Infinite Cache is to improve performance by limiting disk I/O,
you should avoid dedicating so much memory to Infinite Cache that the operating system
must start swapping data to disk. If the operating system begins to swap to disk, you lose
the benefits offered by Infinite Cache.

The overall demand for physical memory can vary throughout the day; if the server is
frequently idle, you may never encounter swapping. If you have dedicated a large
portion of physical memory to the cache, and system usage increases, the operating
system may start swapping. To get the best performance and avoid disk swapping,
dedicate a server node to Infinite Cache so other applications on that computer will not
compete for physical memory.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 285

 EDB Postgres Advanced Server Guide

9 EDB Clone Schema

EDB Clone Schema is an extension module for Advanced Server that allows you to copy
a schema and its database objects from a local or remote database (the source database) to
a receiving database (the target database).

The source and target databases can be the same physical database, or different databases
within the same database cluster, or separate databases running under different database
clusters on separate database server hosts.

Use the following functions with EDB Clone Schema:

localcopyschema. This function makes a copy of a schema and its database

objects from a source database back into the same database (the target), but with a
different schema name than the original. Use this function when the original
source schema and the resulting copy are to reside within the same database. See
Section 9.2.1 for information on the localcopyschema function.
localcopyschema_nb. This function performs the same purpose as
localcopyschema, but as a background job, thus freeing up the terminal from
which the function was initiated. This is referred to as a non-blocking function.
See Section 9.2.2 for information on the localcopyschema_nb function.
remotecopyschema. This function makes a copy of a schema and its database
objects from a source database to a different target database. Use this function
when the original source schema and the resulting copy are to reside in two,
separate databases. The separate databases can reside in the same, or in different
Advanced Server database clusters. See Section 9.2.3 for information on the
remotecopyschema function.
remotecopyschema_nb. This function performs the same purpose as
remotecopyschema, but as a background job, thus freeing up the terminal from
which the function was initiated. This is referred to as a non-blocking function.
See Section 9.2.4 for information on the remotecopyschema_nb function.
process_status_from_log. This function displays the status of the cloning
functions. The information is obtained from a log file that must be specified when
a cloning function is invoked. See Section 9.2.5 for information on the
process_status_from_log function.
remove_log_file_and_job. This function deletes the log file created by a cloning
function. This function can also be used to delete a job created by the non-
blocking form of the function. See Section 9.2.6 for information on the
remove_log_file_and_job function.

The database objects that can be cloned from one schema to another are the following:

Data types
Tables including partitioned tables, but not foreign tables
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 286
 EDB Postgres Advanced Server Guide

Indexes.
Constraints
Sequences
View definitions
Materialized views
Private synonyms
Table triggers, but not event triggers
Rules
Functions
Procedures
Packages
Comments for all supported object types
Access control lists (ACLs) for all supported object types

The following database objects cannot cloned:

Large objects (Postgres LOBs and BFILEs)

Logical replication attributes for a table
Database links
Foreign data wrappers
Foreign tables
Event triggers
Extensions (For cloning objects that rely on extensions, see the third bullet point
in the following limitations list.)
Row level security
Policies
Operator class

In addition, the following limitations apply:

EDB Clone Schema is supported on Advanced Server only when a dialect of

Compatible with Oracle is specified on the Advanced Server Dialect
dialog during installation, or when the --redwood-like keywords are included
during a text mode installation or cluster initialization.

The source code within functions, procedures, triggers, packages, etc., are not
modified after being copied to the target schema. If such programs contain coded
references to objects with schema names, the programs may fail upon invocation
in the target schema if such schema names are no longer consistent within the
target schema.

Cross schema object dependencies are not resolved. If an object in the target
schema depends upon an object in another schema, this dependency is not
resolved by the cloning functions.
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 287
 EDB Postgres Advanced Server Guide

For remote cloning, if an object in the source schema is dependent upon an

extension, then this extension must be created in the public schema of the
remote database before invoking the remote cloning function.

At most, 16 copy jobs can run in parallel to clone schemas, whereas each job can
have at most 16 worker processes to copy table data in parallel.

Queries being run by background workers cannot be cancelled.

The following section describes how to set up EDB Clone Schema on the databases.

9.1 Setup Process

Several extensions along with the PL/Perl language must be installed on any database to
be used as the source or target database by an EDB Clone Schema function.

In addition, some configuration parameters in the postgresql.conf file of the

database servers may benefit from some modification.

The following is the setup instructions for these requirements.

9.1.1 Installing Extensions and PL/Perl

The following describes the steps to install the required extensions and the PL/Perl
language.

These steps must be performed on any database to be used as the source or target
database by an EDB Clone Schema function.

Step 1: The following extensions must be installed on the database:

postgres_fdw
dblink
adminpack
pgagent

Ensure that pgAgent is installed before creating the pgagent extension. You can use
StackBuilder Plus to download and install pgAgent.

The previously listed extensions can be installed by the following commands if they do
not already exist:

CREATE EXTENSION postgres_fdw SCHEMA public;

CREATE EXTENSION dblink SCHEMA public;
CREATE EXTENSION adminpack;

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 288

 EDB Postgres Advanced Server Guide

CREATE EXTENSION pgagent;

For more information about using the CREATE EXTENSION command, see the
PostgreSQL core documentation at:

https://www.postgresql.org/docs/10/static/sql-createextension.html

Step 2: Modify the postgresql.conf file.

Modify the postgresql.conf file by adding $libdir/parallel_clone to the

shared_preload_libraries configuration parameter as shown by the following
example:
shared_preload_libraries = '$libdir/dbms_pipe,$libdir/dbms_aq,$libdir/parallel_clone'

Step 3: The Perl Procedural Language (PL/Perl) must be installed on the database and the
CREATE TRUSTED LANGUAGE plperl command must be run.

Run StackBuilder Plus, select and download the EDB Language Pack installer, and
proceed with the installation.

Step 4: Once the installation has been completed, edit the configuration file
plLanguages.config located in the Advanced Server installation directory under
subdirectory etc/sysconfig to point to the directory where PL/Perl was installed.

The configuration file is at the following location:

INSTALLER_DIR/etc/sysconfig/plLanguages.config

INSTALL_DIR represents the installation directory of Advanced Server such as

/opt/edb/as10, for example.

Replace the PERL_INSTALL_PATH variable in the plLanguages.config file with the

PL/Perl directory:

EDB_PERL_VERSION=5.24
EDB_PYTHON_VERSION=3.4
EDB_TCL_VERSION=8.6

EDB_PERL_PATH=PERL_INSTALL_PATH
EDB_PYTHON_PATH=PYTHON_INSTALL_PATH
EDB_TCL_PATH=TCL_INSTALL_PATH

The following is the edited example:

EDB_PERL_VERSION=5.24
EDB_PYTHON_VERSION=3.4
EDB_TCL_VERSION=8.6

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 289

 EDB Postgres Advanced Server Guide

EDB_PERL_PATH=/opt/edb/languagepack-10/Perl-5.24
EDB_PYTHON_PATH=PYTHON_INSTALL_PATH
EDB_TCL_PATH=TCL_INSTALL_PATH

Step 5: Restart Advanced Server.

For RHEL 7/CentOS 7:

systemctl restart edb-as-10

For RHEL 6/CentOS 6:

service edb-as-10 restart

Step 6: Connect to the database as a superuser where PL/Perl was installed and run the
following command:

CREATE TRUSTED LANGUAGE plperl;

For more information about using the CREATE LANGUAGE command, see the
PostgreSQL core documentation at:

https://www.postgresql.org/docs/10/static/sql-createlanguage.html

9.1.2 Setting Configuration Parameters

The following sections describe certain configuration parameters that may need to be
altered in the postgresql.conf file.

9.1.2.1 Performance Configuration Parameters

You may need to tune the system for copying a large schema as part of one transaction.

Tuning of configuration parameters is for the source database server referenced in a

cloning function.

The configuration parameters in the postgresql.conf file that may need to be tuned
include the following:

work_mem. Specifies the amount of memory to be used by internal sort

operations and hash tables before writing to temporary disk files.
maintenance_work_mem. Specifies the maximum amount of memory to be used
by maintenance operations, such as VACUUM, CREATE INDEX, and ALTER
TABLE ADD FOREIGN KEY.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 290

 EDB Postgres Advanced Server Guide

max_worker_processes. Sets the maximum number of background processes

that the system can support.
checkpoint_timeout. Maximum time between automatic WAL checkpoints, in
seconds.
checkpoint_completion_target. Specifies the target of checkpoint completion, as
a fraction of total time between checkpoints.
checkpoint_flush_after. Whenever more than checkpoint_flush_after
bytes have been written while performing a checkpoint, attempt to force the OS
to issue these writes to the underlying storage.
max_wal_size. Maximum size to let the WAL grow to between automatic WAL
checkpoints.
max_locks_per_transaction. This parameter controls the average number of
object locks allocated for each transaction; individual transactions can lock more
objects as long as the locks of all transactions fit in the lock table.

For information about the configuration parameters, see the PostgreSQL core
documentation at:

https://www.postgresql.org/docs/10/static/runtime-config.html

9.1.2.2 Status Logging

Status logging by the cloning functions creates log files in the directory specified by the
log_directory parameter in the postgresql.conf file for the database server to
which you are connected when invoking the cloning function.

The default location is PGDATA/log as shown by the following:

#log_directory = 'log' # directory where log files are written,

# can be absolute or relative to PGDATA

This directory must exist prior to running a cloning function.

The name of the log file is determined by what you specify in the parameter list when
invoking the cloning function.

To display the status from a log file, use the process_status_from_log function as
described in Section 9.2.5.

To delete a log file, use the remove_log_file_and_job function as described in

Section 9.2.6, or simply navigate to the log directory and delete it manually.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 291

 EDB Postgres Advanced Server Guide

9.1.3 Installing EDB Clone Schema

The following are the directions for installing EDB Clone Schema.

These steps must be performed on any database to be used as the source or target
database by an EDB Clone Schema function.

Step 1: If you had previously installed an older version of the edb_cloneschema

extension, then you must run the following command:

DROP EXTENSION parallel_clone CASCADE;

This command also drops the edb_cloneschema extension.

Step 2: Install the extensions using the following commands:

CREATE EXTENSION parallel_clone SCHEMA public;

CREATE EXTENSION edb_cloneschema;

Make sure you create the parallel_clone extension before creating the
edb_cloneschema extension.

9.1.4 Creating the Foreign Servers and User Mappings

When using one of the local cloning functions, localcopyschema or

localcopyschema_nb, one of the required parameters includes a single, foreign server
for identifying the database server along with its database that is the source and the
receiver of the cloned schema.

When using one of the remote cloning functions, remotecopyschema or

remotecopyschema_nb, two of the required parameters include two foreign servers.
The foreign server specified as the first parameter identifies the source database server
along with its database that is the provider of the cloned schema. The foreign server
specified as the second parameter identifies the target database server along with its
database that is the receiver of the cloned schema.

For each foreign server, a user mapping must be created. When a selected database
superuser invokes a cloning function, that database superuser who invokes the function
must have been mapped to a database user name and password that has access to the
foreign server that is specified as a parameter in the cloning function.

For general information about foreign data, foreign servers, and user mappings, see the
PostgreSQL core documentation at:

https://www.postgresql.org/docs/10/static/ddl-foreign-data.html

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 292

 EDB Postgres Advanced Server Guide

The following two sections describe how these foreign servers and user mappings are
defined.

9.1.4.1 Foreign Server and User Mapping for Local Cloning Functions

For the localcopyschema and localcopyschema_nb functions, the source and target
schemas are both within the same database of the same database server. Thus, only one
foreign server must be defined and specified for these functions. This foreign server is
also referred to as the local server.

This server is referred to as the local server because this server is the one to which you
must be connected when invoking the localcopyschema or localcopyschema_nb
function.

The user mapping defines the connection and authentication information for the foreign
server.

This foreign server and user mapping must be created within the database of the
local server in which the cloning is to occur.

The database user for whom the user mapping is defined must be a superuser and
the user connected to the local server when invoking an EDB Clone Schema
function.

The following example creates the foreign server for the database containing the schema
to be cloned, and to receive the cloned schema as well.

CREATE SERVER local_server FOREIGN DATA WRAPPER postgres_fdw

OPTIONS(
host 'localhost',
port '5444',
dbname 'edb'
);

For more information about using the CREATE SERVER command, see the PostgreSQL
core documentation at:

https://www.postgresql.org/docs/10/static/sql-createserver.html

The user mapping for this server is the following:

CREATE USER MAPPING FOR enterprisedb SERVER local_server

OPTIONS (
user 'enterprisedb',
password 'password'
);

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 293

 EDB Postgres Advanced Server Guide

For more information about using the CREATE USER MAPPING command, see the
PostgreSQL core documentation at:

https://www.postgresql.org/docs/10/static/sql-createusermapping.html

The following psql commands show the foreign server and user mapping:

edb=# \des+
List of foreign servers
-[ RECORD 1 ]--------+----------------------------------------------
Name | local_server
Owner | enterprisedb
Foreign-data wrapper | postgres_fdw
Access privileges |
Type |
Version |
FDW options | (host 'localhost', port '5444', dbname 'edb')
Description |

edb=# \deu+
List of user mappings
Server | User name | FDW options
--------------+--------------+----------------------------------------------
local_server | enterprisedb | ("user" 'enterprisedb', password 'password')
(1 row)

When database superuser enterprisedb invokes a cloning function, the database user
enterprisedb with its password is used to connect to local_server on the
localhost with port 5444 to database edb.

In this case, the mapped database user, enterprisedb, and the database user,
enterprisedb, used to connect to the local edb database happen to be the same,
identical database user, but that is not an absolute requirement.

For specific usage of these foreign server and user mapping examples, see the example
given in Section 9.2.1.

9.1.4.2 Foreign Server and User Mapping for Remote Cloning

Functions

For the remotecopyschema and remotecopyschema_nb functions, the source and

target schemas are in different databases of either the same or different database servers.
Thus, two foreign servers must be defined and specified for these functions.

The foreign server defining the originating database server and its database containing the
source schema to be cloned is referred to as the source server or the remote server.

The foreign server defining the database server and its database to receive the schema to
be cloned is referred to as the target server or the local server.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 294

 EDB Postgres Advanced Server Guide

The target server is also referred to as the local server because this server is the one to
which you must be connected when invoking the remotecopyschema or
remotecopyschema_nb function.

The user mappings define the connection and authentication information for the foreign
servers.

All of these foreign servers and user mappings must be created within the target
database of the target/local server.

The database user for whom the user mappings are defined must be a superuser
and the user connected to the local server when invoking an EDB Clone Schema
function.

The following example creates the foreign server for the local, target database that is to
receive the cloned schema.

CREATE SERVER tgt_server FOREIGN DATA WRAPPER postgres_fdw

OPTIONS(
host 'localhost',
port '5444',
dbname 'tgtdb'
);

The user mapping for this server is the following:

CREATE USER MAPPING FOR enterprisedb SERVER tgt_server

OPTIONS (
user 'tgtuser',
password 'tgtpassword'
);

The following example creates the foreign server for the remote, source database that is
to be the source for the cloned schema.

CREATE SERVER src_server FOREIGN DATA WRAPPER postgres_fdw

OPTIONS(
host '192.168.2.28',
port '5444',
dbname 'srcdb'
);

The user mapping for this server is the following:

CREATE USER MAPPING FOR enterprisedb SERVER src_server

OPTIONS (
user 'srcuser',
password 'srcpassword'
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 295
 EDB Postgres Advanced Server Guide

);

The following psql commands show the foreign servers and user mappings:
tgtdb=# \des+
List of foreign servers
-[ RECORD 1 ]--------+---------------------------------------------------
Name | src_server
Owner | tgtuser
Foreign-data wrapper | postgres_fdw
Access privileges |
Type |
Version |
FDW options | (host '192.168.2.28', port '5444', dbname 'srcdb')
Description |
-[ RECORD 2 ]--------+---------------------------------------------------
Name | tgt_server
Owner | tgtuser
Foreign-data wrapper | postgres_fdw
Access privileges |
Type |
Version |
FDW options | (host 'localhost', port '5444', dbname 'tgtdb')
Description |

tgtdb=# \deu+
List of user mappings
Server | User name | FDW options
------------+--------------+--------------------------------------------
src_server | enterprisedb | ("user" 'srcuser', password 'srcpassword')
tgt_server | enterprisedb | ("user" 'tgtuser', password 'tgtpassword')
(2 rows)

When database superuser enterprisedb invokes a cloning function, the database user
tgtuser with password tgtpassword is used to connect to tgt_server on the
localhost with port 5444 to database tgtdb.

In addition, database user srcuser with password srcpassword connects to

src_server on host 192.168.2.28 with port 5444 to database srcdb.

Note: Be sure the pg_hba.conf file of the database server running the source database
srcdb has an appropriate entry permitting connection from the target server location
(address 192.168.2.27 in the following example) connecting with the database user
srcuser that was included in the user mapping for the foreign server src_server
defining the source server and database.

# TYPE DATABASE USER ADDRESS METHOD

# "local" is for Unix domain socket connections only

local all all md5
# IPv4 local connections:
host srcdb srcuser 192.168.2.27/32 md5

For specific usage of these foreign server and user mapping examples, see the example
given in Section 9.2.3.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 296

 EDB Postgres Advanced Server Guide

9.2 EDB Clone Schema Functions

The EDB Clone Schema functions are created in the edb_util schema when the
parallel_clone and edb_cloneschema extensions are installed.

Verify the following conditions before using an EDB Clone Schema function:

You are connected to the target or local database as the database superuser
defined in the CREATE USER MAPPING command for the foreign server of the
target or local database. See Section 9.1.4.1 for information on the user mapping
for the localcopyschema or localcopyschema_nb function. See Section
9.1.4.2 for information on the user mapping for the remotecopyschema or
remotecopyschema_nb function.
The edb_util schema is in the search path, or the cloning function is to be
invoked with the edb_util prefix.
The target schema does not exist in the target database.
When using the remote copy functions, if the on_tblspace parameter is to be
set to true, then the target database cluster contains all tablespaces that are
referenced by objects in the source schema, otherwise creation of the DDL
statements for those database objects will fail in the target schema. This causes a
failure of the cloning process.
When using the remote copy functions, if the copy_acls parameter is to be set
to true, then all roles that have GRANT privileges on objects in the source schema
exist in the target database cluster, otherwise granting of privileges to those roles
will fail in the target schema. This causes a failure of the cloning process.
pgAgent is running against the target database if the non-blocking form of the
function is to be used.

For information about pgAgent, see the following section of the pgAdmin documentation
available at:

https://www.pgadmin.org/docs/pgadmin4/dev/pgagent.html

Note that pgAgent is provided as a component with Advanced Server.

9.2.1 localcopyschema

The localcopyschema function copies a schema and its database objects within a local
database specified within the source_fdw foreign server from the source schema to the
specified target schema within the same database.

localcopyschema(
source_fdw TEXT,
source_schema TEXT,
target_schema TEXT,
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 297
 EDB Postgres Advanced Server Guide

log_filename TEXT
[, on_tblspace BOOLEAN
[, verbose_on BOOLEAN
[, copy_acls BOOLEAN
[, worker_count INTEGER ]]]]
)

A BOOLEAN value is returned by the function. If the function succeeds, then true is
returned. If the function fails, then false is returned.

The source_fdw, source_schema, target_schema, and log_filename are

required parameters while all other parameters are optional.

Parameters

source_fdw

Name of the foreign server managed by the postgres_fdw foreign data wrapper
from which database objects are to be cloned.

source_schema

Name of the schema from which database objects are to be cloned.

target_schema

Name of the schema into which database objects are to be cloned from the source
schema.

log_filename

Name of the log file in which information from the function is recorded. The log
file is created under the directory specified by the log_directory configuration
parameter in the postgresql.conf file.

on_tblspace

BOOLEAN value to specify whether or not database objects are to be created within
their tablespaces. If false is specified, then the TABLESPACE clause is not
included in the applicable CREATE DDL statement when added to the target
schema. If true is specified, then the TABLESPACE clause is included in the
CREATE DDL statement when added to the target schema. If the on_tblspace
parameter is omitted, the default value is false.

verbose_on

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 298

 EDB Postgres Advanced Server Guide

BOOLEAN value to specify whether or not the DDLs are to be printed in

log_filename when creating objects in the target schema. If false is
specified, then DDLs are not printed. If true is specified, then DDLs are printed.
If omitted, the default value is false.

copy_acls

BOOLEAN value to specify whether or not the access control list (ACL) is to be
included while creating objects in the target schema. The access control list is the
set of GRANT privilege statements. If false is specified, then the access control
list is not included for the target schema. If true is specified, then the access
control list is included for the target schema. If the copy_acls parameter is
omitted, the default value is false.

worker_count

Number of background workers to perform the clone in parallel. If omitted, the

default value is 1.

Example

The following example shows the cloning of schema edb containing a set of database
objects to target schema edbcopy, both within database edb as defined by
local_server.

The example is for the following environment:

Host on which the database server is running: localhost

Port of the database server: 5444
Database source/target of the clone: edb
Foreign server (local_server) and user mapping (see Section 9.1.4.1) with the
information of the preceding bullet points
Source schema: edb
Target schema: edbcopy
Database superuser to invoke localcopyschema: enterprisedb

Before invoking the function, the connection is made by database user enterprisedb
to database edb.

edb=# SET search_path TO "$user",public,edb_util;

SET
edb=# SHOW search_path;
search_path
---------------------------
"$user", public, edb_util
(1 row)

edb=# SELECT localcopyschema ('local_server','edb','edbcopy','clone_edb_edbcopy');

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 299
 EDB Postgres Advanced Server Guide
localcopyschema
-----------------
t
(1 row)

The following displays the logging status using the process_status_from_log

function:
edb=# SELECT process_status_from_log('clone_edb_edbcopy');
process_status_from_log
----------------------------------------------------------------------------------------------
--
(FINISH,"2017-06-29 11:07:36.830783-04",3855,INFO,"STAGE: FINAL","successfully cloned
schema")
(1 row)

After the clone has completed, the following shows some of the database objects copied
to the edbcopy schema:

edb=# SET search_path TO edbcopy;

SET
edb=# \dt+
List of relations
Schema | Name | Type | Owner | Size | Description
---------+---------+-------+--------------+------------+-------------
edbcopy | dept | table | enterprisedb | 8192 bytes |
edbcopy | emp | table | enterprisedb | 8192 bytes |
edbcopy | jobhist | table | enterprisedb | 8192 bytes |
(3 rows)

edb=# \dv
List of relations
Schema | Name | Type | Owner
---------+----------+------+--------------
edbcopy | salesemp | view | enterprisedb
(1 row)

edb=# \di
List of relations
Schema | Name | Type | Owner | Table
---------+---------------+-------+--------------+---------
edbcopy | dept_dname_uq | index | enterprisedb | dept
edbcopy | dept_pk | index | enterprisedb | dept
edbcopy | emp_pk | index | enterprisedb | emp
edbcopy | jobhist_pk | index | enterprisedb | jobhist
(4 rows)

edb=# \ds
List of relations
Schema | Name | Type | Owner
---------+------------+----------+--------------
edbcopy | next_empno | sequence | enterprisedb
(1 row)

edb=# SELECT DISTINCT schema_name, name, type FROM user_source WHERE

schema_name = 'EDBCOPY' ORDER BY type, name;
schema_name | name | type
-------------+--------------------------------+--------------
EDBCOPY | EMP_COMP | FUNCTION
EDBCOPY | HIRE_CLERK | FUNCTION
EDBCOPY | HIRE_SALESMAN | FUNCTION
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 300
 EDB Postgres Advanced Server Guide

EDBCOPY | NEW_EMPNO | FUNCTION

EDBCOPY | EMP_ADMIN | PACKAGE
EDBCOPY | EMP_ADMIN | PACKAGE BODY
EDBCOPY | EMP_QUERY | PROCEDURE
EDBCOPY | EMP_QUERY_CALLER | PROCEDURE
EDBCOPY | LIST_EMP | PROCEDURE
EDBCOPY | SELECT_EMP | PROCEDURE
EDBCOPY | EMP_SAL_TRIG | TRIGGER
EDBCOPY | "RI_ConstraintTrigger_a_19991" | TRIGGER
EDBCOPY | "RI_ConstraintTrigger_a_19992" | TRIGGER
EDBCOPY | "RI_ConstraintTrigger_a_19999" | TRIGGER
EDBCOPY | "RI_ConstraintTrigger_a_20000" | TRIGGER
EDBCOPY | "RI_ConstraintTrigger_a_20004" | TRIGGER
EDBCOPY | "RI_ConstraintTrigger_a_20005" | TRIGGER
EDBCOPY | "RI_ConstraintTrigger_c_19993" | TRIGGER
EDBCOPY | "RI_ConstraintTrigger_c_19994" | TRIGGER
EDBCOPY | "RI_ConstraintTrigger_c_20001" | TRIGGER
EDBCOPY | "RI_ConstraintTrigger_c_20002" | TRIGGER
EDBCOPY | "RI_ConstraintTrigger_c_20006" | TRIGGER
EDBCOPY | "RI_ConstraintTrigger_c_20007" | TRIGGER
EDBCOPY | USER_AUDIT_TRIG | TRIGGER
(24 rows)

9.2.2 localcopyschema_nb

The localcopyschema_nb function copies a schema and its database objects within a
local database specified within the source_fdw foreign server from the source schema
to the specified target schema within the same database, but in a non-blocking manner as
a job submitted to pgAgent.

localcopyschema_nb(
source_fdw TEXT,
source TEXT,
target TEXT,
log_filename TEXT
[, on_tblspace BOOLEAN
[, verbose_on BOOLEAN
[, copy_acls BOOLEAN
[, worker_count INTEGER ]]]]
)

An INTEGER value job ID is returned by the function for the job submitted to pgAgent. If
the function fails, then null is returned.

The source_fdw, source, target, and log_filename are required parameters while
all other parameters are optional.

After completion of the pgAgent job, remove the job with the
remove_log_file_and_job function (see Section 9.2.6).

Parameters

source_fdw
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 301
 EDB Postgres Advanced Server Guide

Name of the foreign server managed by the postgres_fdw foreign data wrapper
from which database objects are to be cloned.

source

Name of the schema from which database objects are to be cloned.

target

Name of the schema into which database objects are to be cloned from the source
schema.

log_filename

Name of the log file in which information from the function is recorded. The log
file is created under the directory specified by the log_directory configuration
parameter in the postgresql.conf file.

on_tblspace

BOOLEAN value to specify whether or not database objects are to be created within
their tablespaces. If false is specified, then the TABLESPACE clause is not
included in the applicable CREATE DDL statement when added to the target
schema. If true is specified, then the TABLESPACE clause is included in the
CREATE DDL statement when added to the target schema. If the on_tblspace
parameter is omitted, the default value is false.

verbose_on

BOOLEAN value to specify whether or not the DDLs are to be printed in

log_filename when creating objects in the target schema. If false is
specified, then DDLs are not printed. If true is specified, then DDLs are printed.
If omitted, the default value is false.

copy_acls

BOOLEAN value to specify whether or not the access control list (ACL) is to be
included while creating objects in the target schema. The access control list is the
set of GRANT privilege statements. If false is specified, then the access control
list is not included for the target schema. If true is specified, then the access
control list is included for the target schema. If the copy_acls parameter is
omitted, the default value is false.

worker_count

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 302

 EDB Postgres Advanced Server Guide

Number of background workers to perform the clone in parallel. If omitted, the

default value is 1.

Example

The same cloning operation is performed as the example in Section 9.2.1, but using the
non-blocking function localcopyschema_nb.

The following command can be used to observe if pgAgent is running on the appropriate
local database:
[root@localhost ~]# ps -ef | grep pgagent
root 4518 1 0 11:35 pts/1 00:00:00 pgagent -s /tmp/pgagent_edb_log
hostaddr=127.0.0.1 port=5444 dbname=edb user=enterprisedb password=password
root 4525 4399 0 11:35 pts/1 00:00:00 grep --color=auto pgagent

If pgAgent is not running, it can be started as shown by the following. The pgagent
program file is located in the bin subdirectory of the Advanced Server installation
directory.
[root@localhost bin]# ./pgagent -l 2 -s /tmp/pgagent_edb_log hostaddr=127.0.0.1 port=5444
dbname=edb user=enterprisedb password=password

Note: the pgagent -l 2 option starts pgAgent in DEBUG mode, which logs continuous
debugging information into the log file specified with the -s option. Use a lower value
for the -l option, or omit it entirely to record less information.

The localcopyschema_nb function returns the job ID shown as 4 in the example.

edb=# SELECT edb_util.localcopyschema_nb ('local_server','edb','edbcopy','clone_edb_edbcopy');
localcopyschema_nb
--------------------
4
(1 row)

The following displays the job status:

edb=# SELECT edb_util.process_status_from_log('clone_edb_edbcopy');
process_status_from_log
----------------------------------------------------------------------------------------------
-----
(FINISH,"29-JUN-17 11:39:11.620093 -04:00",4618,INFO,"STAGE: FINAL","successfully cloned
schema")
(1 row)

The following removes the pgAgent job:

edb=# SELECT edb_util.remove_log_file_and_job (4);
remove_log_file_and_job
-------------------------
t
(1 row)

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 303

 EDB Postgres Advanced Server Guide

9.2.3 remotecopyschema

The remotecopyschema function copies a schema and its database objects from a
source schema in the remote source database specified within the source_fdw foreign
server to a target schema in the local target database specified within the target_fdw
foreign server.

remotecopyschema(
source_fdw TEXT,
target_fdw TEXT,
source_schema TEXT,
target_schema TEXT,
log_filename TEXT
[, on_tblspace BOOLEAN
[, verbose_on BOOLEAN
[, copy_acls BOOLEAN
[, worker_count INTEGER ]]]]
)

A BOOLEAN value is returned by the function. If the function succeeds, then true is
returned. If the function fails, then false is returned.

The source_fdw, target_fdw, source_schema, target_schema, and

log_filename are required parameters while all other parameters are optional.

Parameters

source_fdw

Name of the foreign server managed by the postgres_fdw foreign data wrapper
from which database objects are to be cloned.

target_fdw

Name of the foreign server managed by the postgres_fdw foreign data wrapper
to which database objects are to be cloned.

source_schema

Name of the schema from which database objects are to be cloned.

target_schema

Name of the schema into which database objects are to be cloned from the source
schema.

log_filename
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 304
 EDB Postgres Advanced Server Guide

Name of the log file in which information from the function is recorded. The log
file is created under the directory specified by the log_directory configuration
parameter in the postgresql.conf file.

on_tblspace

BOOLEAN value to specify whether or not database objects are to be created within
their tablespaces. If false is specified, then the TABLESPACE clause is not
included in the applicable CREATE DDL statement when added to the target
schema. If true is specified, then the TABLESPACE clause is included in the
CREATE DDL statement when added to the target schema. If the on_tblspace
parameter is omitted, the default value is false.

Note: If true is specified and a database object has a TABLESPACE clause, but
that tablespace does not exist in the target database cluster, then the cloning
function fails.

verbose_on

BOOLEAN value to specify whether or not the DDLs are to be printed in

log_filename when creating objects in the target schema. If false is
specified, then DDLs are not printed. If true is specified, then DDLs are printed.
If omitted, the default value is false.

copy_acls

BOOLEAN value to specify whether or not the access control list (ACL) is to be
included while creating objects in the target schema. The access control list is the
set of GRANT privilege statements. If false is specified, then the access control
list is not included for the target schema. If true is specified, then the access
control list is included for the target schema. If the copy_acls parameter is
omitted, the default value is false.

Note: If true is specified and a role with GRANT privilege does not exist in the
target database cluster, then the cloning function fails.

worker_count

Number of background workers to perform the clone in parallel. If omitted, the

default value is 1.

Example

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 305

 EDB Postgres Advanced Server Guide

The following example shows the cloning of schema srcschema within database srcdb
as defined by src_server to target schema tgtschema within database tgtdb as
defined by tgt_server.

The source server environment:

Host on which the source database server is running: 192.168.2.28

Port of the source database server: 5444
Database source of the clone: srcdb
Foreign server (src_server) and user mapping (see Section 9.1.4.2) with the
information of the preceding bullet points
Source schema: srcschema

The target server environment:

Host on which the target database server is running: localhost

Port of the target database server: 5444
Database target of the clone: tgtdb
Foreign server (tgt_server) and user mapping (see Section 9.1.4.2) with the
information of the preceding bullet points
Target schema: tgtschema
Database superuser to invoke remotecopyschema: enterprisedb

Before invoking the function, the connection is made by database user enterprisedb
to database tgtdb. A worker_count of 4 is specified for this function.
tgtdb=# SELECT edb_util.remotecopyschema
('src_server','tgt_server','srcschema','tgtschema','clone_rmt_src_tgt',worker_count => 4);
remotecopyschema
------------------
t
(1 row)

The following displays the status from the log file during various points in the cloning
process:
tgtdb=# SELECT edb_util.process_status_from_log('clone_rmt_src_tgt');
process_status_from_log

----------------------------------------------------------------------------------------------
-------------------------------------------
---
(RUNNING,"28-JUN-17 13:18:05.299953 -04:00",4021,INFO,"STAGE: DATA-COPY","[0][0] successfully
copied data in [tgtschema.pgbench_tellers]
")
(1 row)

tgtdb=# SELECT edb_util.process_status_from_log('clone_rmt_src_tgt');

process_status_from_log

----------------------------------------------------------------------------------------------
-------------------------------------------

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 306

 EDB Postgres Advanced Server Guide
---
(RUNNING,"28-JUN-17 13:18:06.634364 -04:00",4022,INFO,"STAGE: DATA-COPY","[0][1] successfully
copied data in [tgtschema.pgbench_history]
")
(1 row)

tgtdb=# SELECT edb_util.process_status_from_log('clone_rmt_src_tgt');

process_status_from_log

----------------------------------------------------------------------------------------------
-------------------------------------------
--
(RUNNING,"28-JUN-17 13:18:10.550393 -04:00",4039,INFO,"STAGE: POST-DATA","CREATE PRIMARY KEY
CONSTRAINT pgbench_tellers_pkey successful"
)
(1 row)

tgtdb=# SELECT edb_util.process_status_from_log('clone_rmt_src_tgt');

process_status_from_log
----------------------------------------------------------------------------------------------
-------------------
(FINISH,"28-JUN-17 13:18:12.019627 -04:00",4039,INFO,"STAGE: FINAL","successfully clone
schema into tgtschema")
(1 row)

The following shows the cloned tables:

tgtdb=# \dt+
List of relations
Schema | Name | Type | Owner | Size | Description
-----------+------------------+-------+--------------+------------+-------------
tgtschema | pgbench_accounts | table | enterprisedb | 256 MB |
tgtschema | pgbench_branches | table | enterprisedb | 8192 bytes |
tgtschema | pgbench_history | table | enterprisedb | 25 MB |
tgtschema | pgbench_tellers | table | enterprisedb | 16 kB |
(4 rows)

When the remotecopyschema function was invoked, four background workers were
specified.

The following portion of the log file clone_rmt_src_tgt shows the status of the
parallel data copying operation using four background workers:
Wed Jun 28 13:18:05.232949 2017 EDT: [4019] INFO: [STAGE: DATA-COPY] [0] table count [4]
Wed Jun 28 13:18:05.233321 2017 EDT: [4019] INFO: [STAGE: DATA-COPY] [0][0] worker started to
copy data
Wed Jun 28 13:18:05.233640 2017 EDT: [4019] INFO: [STAGE: DATA-COPY] [0][1] worker started to
copy data
Wed Jun 28 13:18:05.233919 2017 EDT: [4019] INFO: [STAGE: DATA-COPY] [0][2] worker started to
copy data
Wed Jun 28 13:18:05.234231 2017 EDT: [4019] INFO: [STAGE: DATA-COPY] [0][3] worker started to
copy data
Wed Jun 28 13:18:05.298174 2017 EDT: [4024] INFO: [STAGE: DATA-COPY] [0][3] successfully
copied data in [tgtschema.pgbench_branches]
Wed Jun 28 13:18:05.299913 2017 EDT: [4021] INFO: [STAGE: DATA-COPY] [0][0] successfully
copied data in [tgtschema.pgbench_tellers]
Wed Jun 28 13:18:06.634310 2017 EDT: [4022] INFO: [STAGE: DATA-COPY] [0][1] successfully
copied data in [tgtschema.pgbench_history]
Wed Jun 28 13:18:10.477333 2017 EDT: [4023] INFO: [STAGE: DATA-COPY] [0][2] successfully
copied data in [tgtschema.pgbench_accounts]
Wed Jun 28 13:18:10.477609 2017 EDT: [4019] INFO: [STAGE: DATA-COPY] [0] all workers finished
[4]
Wed Jun 28 13:18:10.477654 2017 EDT: [4019] INFO: [STAGE: DATA-COPY] [0] copy done [4] tables
Wed Jun 28 13:18:10.493938 2017 EDT: [4019] INFO: [STAGE: DATA-COPY] successfully copied data
into tgtschema
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 307
 EDB Postgres Advanced Server Guide

Note that the DATA-COPY log message includes two, square bracket numbers (for
example, [0][3]).

The first number is the job index whereas the second number is the worker index. The
worker index values range from 0 to 3 for the four background workers.

In case two clone schema jobs are running in parallel, the first log file will have 0 as the
job index whereas the second will have 1 as the job index.

9.2.4 remotecopyschema_nb

The remotecopyschema_nb function copies a schema and its database objects from a
source schema in the remote source database specified within the source_fdw foreign
server to a target schema in the local target database specified within the target_fdw
foreign server, but in a non-blocking manner as a job submitted to pgAgent.

remotecopyschema_nb(
source_fdw TEXT,
target_fdw TEXT,
source TEXT,
target TEXT,
log_filename TEXT
[, on_tblspace BOOLEAN
[, verbose_on BOOLEAN
[, copy_acls BOOLEAN
[, worker_count INTEGER ]]]]
)

An INTEGER value job ID is returned by the function for the job submitted to pgAgent. If
the function fails, then null is returned.

The source_fdw, target_fdw, source, target, and log_filename are required

parameters while all other parameters are optional.

After completion of the pgAgent job, remove the job with the
remove_log_file_and_job function (see Section 9.2.6).

Parameters

source_fdw

Name of the foreign server managed by the postgres_fdw foreign data wrapper
from which database objects are to be cloned.

target_fdw

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 308

 EDB Postgres Advanced Server Guide

Name of the foreign server managed by the postgres_fdw foreign data wrapper
to which database objects are to be cloned.

source

Name of the schema from which database objects are to be cloned.

target

Name of the schema into which database objects are to be cloned from the source
schema.

log_filename

Name of the log file in which information from the function is recorded. The log
file is created under the directory specified by the log_directory configuration
parameter in the postgresql.conf file.

on_tblspace

BOOLEAN value to specify whether or not database objects are to be created within
their tablespaces. If false is specified, then the TABLESPACE clause is not
included in the applicable CREATE DDL statement when added to the target
schema. If true is specified, then the TABLESPACE clause is included in the
CREATE DDL statement when added to the target schema. If the on_tblspace
parameter is omitted, the default value is false.

Note: If true is specified and a database object has a TABLESPACE clause, but
that tablespace does not exist in the target database cluster, then the cloning
function fails.

verbose_on

BOOLEAN value to specify whether or not the DDLs are to be printed in

log_filename when creating objects in the target schema. If false is
specified, then DDLs are not printed. If true is specified, then DDLs are printed.
If omitted, the default value is false.

copy_acls

BOOLEAN value to specify whether or not the access control list (ACL) is to be
included while creating objects in the target schema. The access control list is the
set of GRANT privilege statements. If false is specified, then the access control
list is not included for the target schema. If true is specified, then the access

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 309

 EDB Postgres Advanced Server Guide

control list is included for the target schema. If the copy_acls parameter is
omitted, the default value is false.

Note: If true is specified and a role with GRANT privilege does not exist in the
target database cluster, then the cloning function fails.

worker_count

Number of background workers to perform the clone in parallel. If omitted, the

default value is 1.

Example

The same cloning operation is performed as the example in Section 9.2.3, but using the
non-blocking function remotecopyschema_nb.

The following command starts pgAgent on the target database tgtdb. The pgagent
program file is located in the bin subdirectory of the Advanced Server installation
directory.
[root@localhost bin]# ./pgagent -l 1 -s /tmp/pgagent_tgtdb_log hostaddr=127.0.0.1 port=5444
user=enterprisedb dbname=tgtdb password=password

The remotecopyschema_nb function returns the job ID shown as 2 in the example.

tgtdb=# SELECT edb_util.remotecopyschema_nb
('src_server','tgt_server','srcschema','tgtschema','clone_rmt_src_tgt',worker_count => 4);
remotecopyschema_nb
---------------------
2
(1 row)

The completed status of the job is shown by the following:

tgtdb=# SELECT edb_util.process_status_from_log('clone_rmt_src_tgt');
process_status_from_log
----------------------------------------------------------------------------------------------
-------------------
(FINISH,"29-JUN-17 13:16:00.100284 -04:00",3849,INFO,"STAGE: FINAL","successfully clone
schema into tgtschema")
(1 row)

The following removes the log file and the pgAgent job:
tgtdb=# SELECT edb_util.remove_log_file_and_job ('clone_rmt_src_tgt',2);
remove_log_file_and_job
-------------------------
t
(1 row)

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 310

 EDB Postgres Advanced Server Guide

9.2.5 process_status_from_log

The process_status_from_log function provides the status of a cloning function

from its log file.

process_status_from_log (
log_file TEXT
)

The function returns the following fields from the log file:

Table 9-1 - Clone Schema Log File

Field Name Description
status Displays either STARTING, RUNNING, FINISH, or FAILED.
execution_time When the command was executed. Displayed in timestamp format.
pid Session process ID in which clone schema is getting called.
level Displays either INFO, ERROR, or SUCCESSFUL.
Displays either STARTUP, INITIAL, DDL-COLLECTION, PRE-DATA,
stage
DATA-COPY, POST-DATA, or FINAL.
message Information respective to each command or failure.

Parameters

log_file

Name of the log file recording the cloning of a schema as specified when the
cloning function was invoked.

Example

The following shows usage of the process_status_from_log function:

edb=# SELECT edb_util.process_status_from_log('clone_edb_edbcopy');

process_status_from_log
----------------------------------------------------------------------------------------------
-----
(FINISH,"26-JUN-17 11:57:03.214458 -04:00",3691,INFO,"STAGE: FINAL","successfully cloned
schema")
(1 row)

9.2.6 remove_log_file_and_job

The remove_log_file_and_job function performs cleanup tasks by removing the log

files created by the schema cloning functions and the jobs created by the non-blocking
functions.

remove_log_file_and_job (
{ log_file TEXT |

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 311

 EDB Postgres Advanced Server Guide

job_id INTEGER |
log_file TEXT, job_id INTEGER
}
)

Values for any or both of the two parameters may be specified when invoking the
remove_log_file_and_job function:

If only log_file is specified, then the function will only remove the log file.
If only job_id is specified, then the function will only remove the job.
If both are specified, then the function will remove the log file and the job.

Parameters

log_file

Name of the log file to be removed.

job_id

Job ID of the job to be removed.

Example

The following examples removes only the log file, given the log filename.

edb=# SELECT edb_util.remove_log_file_and_job ('clone_edb_edbcopy');

remove_log_file_and_job
-------------------------
t
(1 row)

The following example removes only the job, given the job ID.

edb=# SELECT edb_util.remove_log_file_and_job (3);

remove_log_file_and_job
-------------------------
t
(1 row)

The following example removes the log file and the job, given both values:
tgtdb=# SELECT edb_util.remove_log_file_and_job ('clone_rmt_src_tgt',2);
remove_log_file_and_job
-------------------------
t
(1 row)

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 312

 EDB Postgres Advanced Server Guide

10 PL/Java
The PL/Java package provides access to Java stored procedures, triggers, and functions
via the JDBC interface. Unless otherwise noted, the commands and paths noted in the
following section assume that you have performed an installation with the interactive
installer.

Before installing PL/Java for use with a standard Java virtual machine (JVM) on a Linux
system, you must first confirm that a Java runtime environment (version 1.5 or later) is
installed on your system. Installation of a Java development kit also provides a Java
runtime environment.

10.1 Installing PL/Java on Linux

The following steps outline the process of installing PL/Java on a Linux system:

Step 1: Edit the postgresql.conf file located under the data directory of your
Advanced Server installation and add (or modify) the following settings:

pljava.classpath = 'path_to_pljava.jar'
pljava.libjvm_location = 'path_to_libjvm.so'

Where path_to_pljava.jar specifies the location of the pljava.jar file and

path_to_libjvm.so specifies the location of the libjvm.so file.

For example, the following lists the paths for a default installation with Java version 1.8:

pljava.classpath = '/opt/edb/as10/share/pljava/pljava-
1.5.0.jar'
pljava.libjvm_location = '/usr/lib/jvm/java-1.8.0-openjdk-
1.8.0.91-1.b14.el6.x86_64/jre/lib/amd64/server/libjvm.so'

Step 2: Restart the database server.

Step 3: You can use the CREATE EXTENSION command to install PL/Java. To install the
PL/Java extension, login to the database in which you want to install PL/Java with the
psql or pgAdmin client, and invoke the following command:

CREATE EXTENSION pljava;

Step 4: To confirm that PL/Java is installed, invoke the following command:

SELECT * FROM pg_language WHERE lanname LIKE 'java%';

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 313

 EDB Postgres Advanced Server Guide

The edb-psql client displays two rows indicating that java and javau (Java
Untrusted) have been installed in the database.
edb=# SELECT * FROM pg_language WHERE lanname LIKE 'java%';
lanname | lanowner | lanispl | lanpltrusted | lanplcallfoid | laninline | lanvalidator |
lanacl
---------+----------+---------+--------------+---------------+-----------+--------------+-----
--------------------------
java | 10 | t | t | 16462 | 0 | 0 |
{enterprisedb=U/enterprisedb}
javau | 10 | t | f | 16463 | 0 | 0 |
(2 rows)

10.2 Installing PL/Java on Windows

The following steps outline the process of installing PL/Java on a Windows system.

Step 1: Edit the postgresql.conf file and add (or modify) the following settings:

pljava.classpath = 'POSTGRES_INSTALL_HOME\lib\pljava.jar'
pljava.libjvm_location = 'path_to_libjvm.so'

Where POSTGRES_INSTALL_HOME specifies the location of the Advanced Server

installation. For example, the following is the configuration setting for a default
installation:

pljava.classpath = 'C:\Program Files\edb\as10\lib\pljava.jar'

Step 2: Restart the database server.

Step 3: Modify the PATH setting used by the server, adding the following two entries:

%JRE_HOME%\bin;%JRE_HOME%\bin\client

Where JRE_HOME specifies the installation directory of your Java runtime environment.
If you have a Java development kit, substitute the location of $JDK_HOME/jre for
JRE_HOME.

Step 4: Use the Postgres CREATE EXTENSION command to install PL/Java. To run the
installation script, use the psql or pgAdmin client to connect to the database in which you
wish to install PL/Java and invoke the following command:

CREATE EXTENSION pljava;

Step 5: To confirm that PL/Java is installed, invoke the following command:

SELECT * FROM pg_language WHERE lanname LIKE 'java%';

The client will return a result set that includes java and javau (Java Untrusted).
Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 314
 EDB Postgres Advanced Server Guide

10.3 Using PL/Java

To create a PL/Java program, you must first create a Java class that contains at least one
static method, and then you must compile that class into a .class or .jar file. Next,
you declare the Java function within SQL using the CREATE FUNCTION command. The
CREATE FUNCTION command gives a SQL name to the function and associates the
compiled class (and method name) with that function name.

For example, the following CREATE FUNCTION statement creates a function named
getsysprop:

CREATE FUNCTION getsysprop(VARCHAR)

RETURNS VARCHAR
AS 'java.lang.System.getProperty'
LANGUAGE java;

When invoked, getsysprop will execute the getProperty (static) method defined
within the java.lang.System class.

SELECT getsysprop('user.home');

getsysprop
---------------
/opt/edb/as10
(1 row)

The example that follows demonstrates the procedures used to create and install a simple
HelloWorld program:

Step 1: Save the following code sample to a file named HelloWorld.java:

package com.mycompany.helloworld;
public class HelloWorld
{
public static String helloWorld()
{
return "Hello World";
}
}

Step 2: Compile the file.

$ javac HelloWorld.java

And save it in a folder hierarchy as:

com/mycompany/helloworld/HelloWorld.class

Step 3: Create an archive file (a JAR file) named helloworld.jar:

jar cf helloworld.jar com/mycompany/helloworld/HelloWorld.class

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 315
 EDB Postgres Advanced Server Guide

Step 4: Open the edb-psql client, and install the jar file with the following command:

SELECT sqlj.install_jar('file:///file_path/helloworld.jar',
'helloworld', true);

Where file_path is the directory containing the helloworld.jar file. For example,
if the /tmp directory is the file_path:

SELECT sqlj.install_jar('file:///tmp/helloworld.jar', 'helloworld', true);

To confirm that the jar file has been loaded correctly, perform a SELECT statement on the
sqlj.jar_entry and sqlj.jar_repository tables.

edb=# SELECT entryid, entryname FROM sqlj.jar_entry;

entryid | entryname
---------+-------------------------------------------
1 | com/mycompany/helloworld/HelloWorld.class
(1 row)

edb=# SELECT jarid, jarname, jarorigin, jarowner FROM sqlj.jar_repository;

jarid | jarname | jarorigin | jarowner
-------+------------+----------------------------+--------------
4 | helloworld | file:///tmp/helloworld.jar | enterprisedb
(1 row)

Step 5: Set the classpath as:

edb=# SELECT sqlj.set_classpath('public', 'helloworld');

The sqlj.classpath_entry table will now include an entry for the helloworld
class file.

edb=# SELECT * FROM sqlj.classpath_entry;

schemaname | ordinal | jarid
------------+---------+-------
public | 1 | 4
(1 row)

Step 6: Create a function that uses Java to call the static function declared in the jar file:

CREATE OR REPLACE FUNCTION helloworld()

RETURNS "varchar"
AS
'com.mycompany.helloworld.HelloWorld.helloWorld'
LANGUAGE 'java' VOLATILE;

Step 7: Execute the function:

edb=# SELECT * FROM helloworld();

You should see the output:

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 316

 EDB Postgres Advanced Server Guide

helloworld
-------------
Hello World
(1 row)

The official PL/Java distribution is distributed with examples and documentation. For
more information about using PL/Java, see the project page at:

https://github.com/tada/pljava/wiki

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 317

 EDB Postgres Advanced Server Guide

11 Enhanced SQL Features

Advanced Server includes enhanced SQL functionality that provides additional flexibility
and convenience. This chapter discusses some of these additions.

11.1 COMMENT

In addition to commenting on objects supported by the PostgreSQL COMMENT command,

Advanced Server supports comments on additional object types. The complete supported
syntax is:

COMMENT ON
{
AGGREGATE aggregate_name ( aggregate_signature ) |
CAST (source_type AS target_type) |
COLLATION object_name |
COLUMN relation_name.column_name |
CONSTRAINT constraint_name ON table_name |
CONSTRAINT constraint_name ON DOMAIN domain_name |
CONVERSION object_name |
DATABASE object_name |
DOMAIN object_name |
EXTENSION object_name |
EVENT TRIGGER object_name |
FOREIGN DATA WRAPPER object_name |
FOREIGN TABLE object_name |
FUNCTION func_name ([[argmode] [argname] argtype [, ...]])|
INDEX object_name |
LARGE OBJECT large_object_oid |
MATERIALIZED VIEW object_name |
OPERATOR operator_name (left_type, right_type) |
OPERATOR CLASS object_name USING index_method |
OPERATOR FAMILY object_name USING index_method |
PACKAGE object_name
POLICY policy_name ON table_name |
[ PROCEDURAL ] LANGUAGE object_name |
PROCEDURE proc_name [([[argmode] [argname] argtype [, ...]])]
PUBLIC SYNONYM object_name
ROLE object_name |
RULE rule_name ON table_name |
SCHEMA object_name |
SEQUENCE object_name |
SERVER object_name |
TABLE object_name |
TABLESPACE object_name |
TEXT SEARCH CONFIGURATION object_name |
TEXT SEARCH DICTIONARY object_name |
TEXT SEARCH PARSER object_name |

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 318

 EDB Postgres Advanced Server Guide

TEXT SEARCH TEMPLATE object_name |

TRANSFORM FOR type_name LANGUAGE lang_name |
TRIGGER trigger_name ON table_name |
TYPE object_name |
VIEW object_name
} IS 'text'
where aggregate_signature is:
* |
[ argmode ] [ argname ] argtype [ , ... ] |
[ [ argmode ] [ argname ] argtype [ , ... ] ]
ORDER BY [ argmode ] [ argname ] argtype [ , ... ]

Parameters

object_name

The name of the object on which you are commenting.

AGGREGATE aggregate_name (aggregate_signature)

Include the AGGREGATE clause to create a comment about an aggregate.

aggregate_name specifies the name of an aggregate, and
aggregate_signature specifies the associated signature in one of the
following forms:

* |
[ argmode ] [ argname ] argtype [ , ... ] |
[ [ argmode ] [ argname ] argtype [ , ... ] ]
ORDER BY [ argmode ] [ argname ] argtype [ , ... ]

Where argmode is the mode of a function, procedure, or aggregate argument;

argmode may be IN, OUT, INOUT, or VARIADIC. If omitted, the default is IN.

argname is the name of an aggregate argument.

argtype is the data type of an aggregate argument.

CAST (source_type AS target_type)

Include the CAST clause to create a comment about a cast. When creating a
comment about a cast, source_type specifies the source data type of the cast,
and target_type specifies the target data type of the cast.

COLUMN relation_name.column_name

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 319

 EDB Postgres Advanced Server Guide

Include the COLUMN clause to create a comment about a column. column_name

specifies name of the column to which the comment applies. relation_name is
the table, view, composite type, or foreign table in which a column resides.

CONSTRAINT constraint_name ON table_name

CONSTRAINT constraint_name ON DOMAIN domain_name

Include the CONSTRAINT clause to add a comment about a constraint. When

creating a comment about a constraint, constraint_name specifies the name of
the constraint; table_name or domain_name specifies the name of the table or
domain on which the constraint is defined.

FUNCTION func_name ([[argmode] [argname] argtype [, ...]])

Include the FUNCTION clause to add a comment about a function. func_name

specifies the name of the function. argmode specifies the mode of the function;
argmode may be IN, OUT, INOUT, or VARIADIC. If omitted, the default is IN.
argname specifies the name of a function, procedure, or aggregate argument.
argtype specifies the data type of a function, procedure, or aggregate argument.

large_object_oid

large_object_oid is the system-assigned OID of the large object about which

you are commenting.

OPERATOR operator_name (left_type, right_type)

Include the OPERATOR clause to add a comment about an operator.

operator_name specifies the (optionally schema-qualified) name of an operator
on which you are commenting. left_type and right_type are the
(optionally schema-qualified) data type(s) of the operator's arguments.

OPERATOR CLASS object_name USING index_method

Include the OPERATOR CLASS clause to add a comment about an operator class.
object_name specifies the (optionally schema-qualified) name of an operator on
which you are commenting. index_method specifies the associated index
method of the operator class.

OPERATOR FAMILY object_name USING index_method

Include the OPERATOR FAMILY clause to add a comment about an operator

family. object_name specifies the (optionally schema-qualified) name of an
operator family on which you are commenting. index_method specifies the
associated index method of the operator family.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 320

 EDB Postgres Advanced Server Guide

POLICY policy_name ON table_name

Include the POLICY clause to add a comment about a policy. policy_name

specifies the name of the policy, and table_name specifies the table that the
policy is associated with.

PROCEDURE proc_name [([[argmode] [argname] argtype [, ...]])]

Include the PROCEDURE clause to add a comment about a procedure. proc_name

specifies the name of the procedure. argmode specifies the mode of the
procedure; argmode may be IN, OUT, INOUT, or VARIADIC. If omitted, the
default is IN. argname specifies the name of a function, procedure, or aggregate
argument. argtype specifies the data type of a function, procedure, or aggregate
argument.

RULE rule_name ON table_name

Include the RULE clause to specify a COMMENT on a rule. rule_name specifies

the name of the rule, and table_name specifies the name of the table on which
the rule is defined.

TRANSFORM FOR type_name LANGUAGE lang_name |

Include the TRANSFORM FOR clause to specify a COMMENT on a TRANSFORM.

type_name specifies the name of the data type of the transform and lang_name
specifies the name of the language of the transform.

TRIGGER trigger_name ON table_name

Include the TRIGGER clause to specify a COMMENT on a trigger. trigger_name

specifies the name of the trigger, and table_name specifies the name of the table
on which the trigger is defined.

text

The comment, written as a string literal; or NULL to drop the comment.

Notes:

Names of tables, aggregates, collations, conversions, domains, foreign tables, functions,

indexes, operators, operator classes, operator families, packages, procedures, sequences,
text search objects, types, and views can be schema-qualified.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 321

 EDB Postgres Advanced Server Guide

Example:

The following example adds a comment to a table named new_emp:

COMMENT ON TABLE new_emp IS 'This table contains information about new

employees.';

For more information about using the COMMENT command, please see the PostgreSQL
core documentation at:

https://www.postgresql.org/docs/10/static/sql-comment.html

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 322

 EDB Postgres Advanced Server Guide

12 System Catalog Tables

The following system catalog tables contain definitions of database objects. The layout
of the system tables is subject to change; if you are writing an application that depends on
information stored in the system tables, it would be prudent to use an existing catalog
view, or create a catalog view to isolate the application from changes to the system table.

12.1 edb_dir

The edb_dir table contains one row for each alias that points to a directory created with
the CREATE DIRECTORY command. A directory is an alias for a pathname that allows a
user limited access to the host file system.

You can use a directory to fence a user into a specific directory tree within the file
system. For example, the UTL_FILE package offers functions that permit a user to read
and write files and directories in the host file system, but only allows access to paths that
the database administrator has granted access to via a CREATE DIRECTORY command.

Column Type Modifiers Description

dirname "name" not null The name of the alias.
dirowner oid not null The OID of the user that owns the alias.
dirpath text The directory name to which access is granted.
diracl aclitem[] The access control list that determines which users
may access the alias.

12.2 edb_all_resource_groups

The edb_all_resource_groups table contains one row for each resource group
created with the CREATE RESOURCE GROUP command and displays the number of
active processes in each resource group.

Column Type Modifiers Description

group_name "name" The name of the resource group.
active_processes integer Number of currently active processes in
the resource group.
cpu_rate_limit float8 Maximum CPU rate limit for the
resource group. 0 means no limit.
per_process_cpu_rate_li float8 Maximum CPU rate limit per currently
mit active process in the resource group.
dirty_rate_limit float8 Maximum dirty rate limit for a resource
group. 0 means no limit.
per_process_dirty_rate_ float8 Maximum dirty rate limit per currently
limit active process in the resource group.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 323

 EDB Postgres Advanced Server Guide

12.3 edb_password_history

The edb_password_history table contains one row for each password change. The
table is shared across all databases within a cluster.

Column Type References Description

passhistroleid oid pg_authid.oid The ID of a role.
passhistpassword text Role password in md5 encrypted form.
passhistpasswordsetat timestamptz The time the password was set.

12.4 edb_policy

The edb_policy table contains one row for each policy.

Column Type Modifiers Description

policyname name not null The policy name.
policygroup oid not null Currently unused.
policyobject oid not null The OID of the table secured by this policy
(the object_schema plus the object_name).
policykind char not null The kind of object secured by this policy:
'r' for a table
'v' for a view
= for a synonym
Currently always 'r'.
policyproc oid not null The OID of the policy function
(function_schema plus policy_function).
policyinsert boolean not null True if the policy is enforced by INSERT
statements.
policyselect boolean not null True if the policy is enforced by SELECT
statements.
policydelete boolean not null True if the policy is enforced by DELETE
statements.
policyupdate boolean not null True if the policy is enforced by UPDATE
statements.
policyindex boolean not null Currently unused.
policyenabled boolean not null True if the policy is enabled.
policyupdatecheck boolean not null True if rows updated by an UPDATE
statement must satisfy the policy.
policystatic boolean not null Currently unused.
policytype integer not null Currently unused.
policyopts integer not null Currently unused.
policyseccols int2vector not null The column numbers for columns listed in
sec_relevant_cols.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 324

 EDB Postgres Advanced Server Guide

12.5 edb_profile

The edb_profile table stores information about the available profiles. edb_profiles
is shared across all databases within a cluster.

Column Type References Description

oid oid Row identifier (hidden attribute;
must be explicitly selected).
prfname name The name of the profile.
prffailedloginattempts integer The number of failed login attempts
allowed by the profile. -1 indicates
that the value from the default
profile should be used. -2 indicates
no limit on failed login attempts.
prfpasswordlocktime integer The password lock time associated
with the profile (in seconds). -1
indicates that the value from the
default profile should be used. -2
indicates that the account should be
locked permanently.
prfpasswordlifetime integer The password life time associated
with the profile (in seconds). -1
indicates that the value from the
default profile should be used. -2
indicates that the password never
expires.
prfpasswordgracetime integer The password grace time associated
with the profile (in seconds). -1
indicates that the value from the
default profile should be used. -2
indicates that the password never
expires.
prfpasswordreusetime integer The number of seconds that a user
must wait before reusing a
password. -1 indicates that the
value from the default profile
should be used. -2 indicates that the
old passwords can never be reused.
prfpasswordreusemax integer The number of password changes
that have to occur before a
password can be reused. -1
indicates that the value from the
default profile should be used. -2
indicates that the old passwords can
never be reused.
prfpasswordverifyfuncdb oid pg_database.oid The OID of the database in which
the password verify function exists.
prfpasswordverifyfunc oid pg_proc.oid The OID of the password verify
function associated with the profile.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 325

 EDB Postgres Advanced Server Guide

12.6 edb_resource_group

The edb_resource_group table contains one row for each resource group created with
the CREATE RESOURCE GROUP command.

Column Type Modifiers Description

rgrpname "name" not null The name of the resource group.
rgrpcpuratelimit float8 not null Maximum CPU rate limit for a resource
group. 0 means no limit.
rgrpdirtyratelimit float8 not null Maximum dirty rate limit for a resource
group. 0 means no limit.

12.7 edb_variable

The edb_variable table contains one row for each package level variable (each
variable declared within a package).

Column Type Modifiers Description

varname "name" not null The name of the variable.
varpackage oid not null The OID of the pg_namespace row that stores the
package.
vartype oid not null The OID of the pg_type row that defines the type of
the variable.
varaccess "char" not null + if the variable is visible outside of the package.
- if the variable is only visible within the package.
Note: Public variables are declared within the
package header; private variables are declared
within the package body.
varsrc text Contains the source of the variable declaration,
including any default value expressions for the
variable.
varseq smallint not null The order in which the variable was declared in the
package.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 326

 EDB Postgres Advanced Server Guide

12.8 pg_synonym

The pg_synonym table contains one row for each synonym created with the CREATE
SYNONYM command or CREATE PUBLIC SYNONYM command.

Column Type Modifiers Description

synname "name" not null The name of the synonym.
synnamespace oid not null Replaces synowner. Contains the OID of the
pg_namespace row where the synonym is stored
synowner oid not null The OID of the user that owns the synonym.
synobjschema "name" not null The schema in which the referenced object is
defined.
synobjname "name" not null The name of the referenced object.
synlink text The (optional) name of the database link in which
the referenced object is defined.

12.9 product_component_version

The product_component_version table contains information about feature

compatibility; an application can query this table at installation or run time to verify that
features used by the application are available with this deployment.

Column Type Description

product character varying (74) The name of the product.
version character varying (74) The version number of the product.
status character varying (74) The status of the release.

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 327

 EDB Postgres Advanced Server Guide

13 Advanced Server Keywords

A keyword is a word that is recognized by the Advanced Server parser as having a
special meaning or association. You can use the pg_get_keywords() function to
retrieve an up-to-date list of the Advanced Server keywords:

acctg=#
acctg=# SELECT * FROM pg_get_keywords();
word | catcode | catdesc
---------------------+---------+---------------------------------
abort | U | unreserved
absolute | U | unreserved
access | U | unreserved
...

pg_get_keywords returns a table containing the keywords recognized by Advanced

Server:

The word column displays the keyword.

The catcode column displays a category code.
The catdesc column displays a brief description of the category to which the
keyword belongs.

Note that any character can be used in an identifier if the name is enclosed in double
quotes. You can selectively query the pg_get_keywords() function to retrieve an up-
to-date list of the Advanced Server keywords that belong to a specific category:

SELECT * FROM pg_get_keywords() WHERE catcode = 'code';

Where code is:

R - The word is reserved. Reserved keywords may never be used as an identifier;

they are reserved for use by the server.

U - The word is unreserved. Unreserved words are used internally in some

contexts, but may be used as a name for a database object.

T - The word is used internally, but may be used as a name for a function or type.

C - The word is used internally, and may not be used as a name for a function or
type.

For more information about Advanced Server identifiers and keywords, please refer to
the PostgreSQL core documentation at:

https://www.postgresql.org/docs/10/static/sql-syntax-lexical.html

Copyright 2014 - 2017 EnterpriseDB Corporation. All rights reserved. 328