Workflow Performance Tuning in Release 12.1

Workflow Performance Tuning
in Release 12
Karen Brownfield
Copyright © 2014 Karen Brownfield All Rights Reserved . Any other commercial product names herein are trademark, registered trademarks or service marks of their respective owners.
About the Speaker
• Oracle Ace
• Oracle Certified Specialist (EBS and Fusion)
• Over 20 years E-Business Suite support
• 14 years Oracle Workflow design and support
• OAUG Board 1994-2009, 2014-2015, former President
• Member ATG Customer Advisory Board
• Co-Chair Oracle EBS User Management SIG
• Over 100 presentations worldwide
• Co-Author multiple books on E-Business Suite
– The ABCs of Oracle Workflow for E-Business Suite
Release 11i and Release 12
– The Release 12 Primer – Shining a Light on the
2 Release 12 World
Copyright © 2014 Karen Brownfield All Rights Reserved . Any other commercial product names herein are trademark, registered trademarks or service marks of their respective owners.
• Established in 2001
• SBA 8(a) Small Business
disadvantaged company
• GSA Schedule contract
GS-35F-0680V
• Texas State HUB vendor
• For more information, go
to our web site at
www.Infosemantics.com
– R12.1.3, R12.2, OBIEE
public vision instances
– Posted presentations on
functional and technical
topics
3 Copyright © 2014 Karen Brownfield All Rights Reserved . Any other commercial product names herein are trademark, registered trademarks or service marks of their respective owners.
Agenda
• Introduction
• Patch Current
• Clean Up Old and Errored
• Purge
• Background Processing
• Queue Performance
• Mailer Performance
• Miscellaneous Performance Aides
Which Are You?
The opposite of the ostrich is the rooster who is alert and awake
early to see what is on the horizon.
Rather than fear, he crows loudly a warning to be heeded by all.
Source: http://users.cybertime.net/~ajgood/ostrich.html
Workflow Analyzer
• 1369938.1 “Workflow Analyzer script for E-Business

Suite Workflow Monitoring and Maintenance”
– Better than R12 Workflow Health Check Diagnostic or
11i Workflow Status and Purgeable Items or Workflow
Performance
– SQL Script – Updated frequently
– Run as concurrent request – note 1425053.1
– FAQ – note 1452224.1
– Focuses on Administration and Performance
– https://blogs.oracle.com/oracleworkflow/entry/e_busine
ss_suite_proactive_support
Workflow Analyzer
Workflow Community
• Go to the Oracle
Support
Community tab
• Enter “Workflow”
to go to the
workflow sub-
space
Workflow Community
• Sign up for email notifications
Patch Current
Patch Current
• It’s not just the RUPs, one-offs are important

• Workflow is dependant on HR, AME
– Diagnostics are important also
• Product workflow fixes are provided by product team,
not ATG patches
• Best sources of Current Patches
– Patch Wizard – look for owf
• Workflow now release quarterly rollups
– Workflow Analyzer
– OAUG Workflow SIG
• Updated approximately once/year
• Applied status dependent on being current with above
sources
From Patch Wizard
From Workflow Analyzer
Mailer Patches
Non-Mailer Patches
Clean Up Errors and
Old Workflows
Despite low overall

count and low % closed,
status is red due to age
of closed
Clean up Errors
• Perform following query

SELECT COUNT (*)
,item_type
,activity_name
,MIN (item_begin_date)
,MAX (item_begin_date)
FROM wf_item_activity_statuses_v
WHERE activity_status_code = 'ERROR'
AND item_end_date IS NULL
GROUP BY item_type
,activity_name
ORDER BY 1 DESC, 5 DESC, 2;
• Play with Order By – try 5, 1, 2 also

Clean up Errors
• Triage – Most Recent, Highest Numbers

• It isn’t enough to clean up the errored workflows…
Clean up Associated Error
Item Types
SELECT item_type
,parent_item_type
,DECODE (end_date, NULL, 'OPEN', 'CLOSED')
error_type_status
,COUNT (*)
FROM wf_items
WHERE parent_item_type is not null
AND item_type in ('CUNNLWF','DOSFLOW','DOSFLOWE',
'ECXERROR','HRSSA','HRSTAND','HXCEMP','IBUHPSUB','OKLAMERR',
'OMERROR','PARMAAP','PARMATRX','POERROR','WFSTD','XDPWFSTD',
'ZPBWFERR', 'WFERROR')
GROUP BY item_type
,parent_item_type
,DECODE (end_date, NULL, 'OPEN', 'CLOSED')
ORDER BY 4 DESC,item_type,parent_item_type;
Clean up Associated Error
Item Types
• Purge now closes WFERROR where calling activity is

closed
– WFERROR not the only ERROR Item Type
– Can’t purge if children open
• Notice chains OEOH→OMERROR→WFERROR
OEOH→OEOL→WFERROR
Clean up Event Errors

SELECT COUNT (*)
,v.text_value
,min(i.begin_date)
,max(i.begin_date)
FROM wf_item_attribute_values v
,wf_items i
WHERE v.item_key=i.item_key
AND v.item_type = i.item_type
AND v.item_type = 'WFERROR' The text value of the
AND v.NAME = 'EVENT_NAME' attribute “EVENT
AND v.text_value IS NOT NULL NAME” is the name
of the event
GROUP BY text_value
ORDER BY 4 DESC,1 DESC,text_value;
Clean Up Event Errors
• Find and fix what causes event to error

– See the worklist flexfields section in “Workflow Troubleshooting
in Release 12” to get details on event errors shown below
Clean Up Event Errors
• Message to SYSADMIN can re-raise event if still needs

processing, else abort WFERROR
– Note: after 24 hours, no record of event anywhere
else
Clean up Event Errors
(showing portion of script)
Adjust each
update
similarly
• Adjust $FND_TOP/sql/wfrmtype.sql script

• Always test in non-production instance first
Old Items
Click the link to the sql

script; modify as needed
for more details such as
item types
Product Specific Workflows
HR added v4.09, PO v5.01

Inv, AP Coming soon
Configure (Setup)
Seeded Workflows
• Read the documentation
– Setup
– How the workflow behaves
– My Oracle Support white papers, notes
• Setup not just Builder
– Profile Options
– Approvals Management Engine (AME)
– Hierarchies
– Other Screens
• Account Generator Top Process
• PO Documents – identify workflow to run by document
• GL ledgers page – are you using approvals
Purge
Purge!!!
11i Purgeable for

PERM always 0
• Need schedule for Temporary and for Permanent

• If Purgeable=0, ensure child/parent workflows closed
Purge Obsolete Workflow
Runtime Data
• Schedule Nightly or at minimum Weekly
• Parameters
– Leave Item Type/Item Key blank
– Age – recommended at least 7, no more than 60
– Persistence Type
• One Schedule Temporary, one Permanent
– Core Workflow Only – Set to Y
• At least monthly, run w/ value set to N
– Commit Frequency – leave at default – 500 (that’s 500
workflows, not 500 records)
– Signed Notifications – Customer Choice
– (R12.2) Other Cached Data
Purge – What Happens
• Aborts WFERROR where PARENT_ITEM_TYPE

matches Item Type parameter and where linked activity
(PARENT_CONTEXT) no longer in error status
– But not POERROR, OMERROR or other error types
– Remember…set “Attempt to Close” = Y in the ”Purge
Order Management Workflows” concurrent program to
try to close OMERROR/WFERROR
• Purges Item Types matching Item Type parameter if
END_DATE is not NULL and not linked to open parent or
child workflow
• (R12.2) Other Cached Data = Y

– Purges records from WF_ATTRIBUTE_CACHE
• When user record is created/updated, stores 16+ records of data
about the user (name, email, mailer preference, description, etc)
– Does not store old/new values
– If table purged, does not rebuild when user signs in
• If using OID, contains record when OID sync event is raised
(MOS note 380487.1)
• Used as intermediate table for LDAP sync (MOS note 150832.1)
– MOS note 846989.1 “How to Purge the
Wf_attribute_cache table?” recommends periodically
removing obsoleted data but doesn’t define “obsolete”
• Hint: To purge just WF_ATTRIBUTE_CACHE, use
item_type System: Mailer (as will never be in runtime
history)
• If “Core Workflow Only” = N

– Purges WF_ACTIVITIES table where END_DATE is
not NULL and ACTIVITY_ID is not referenced in active
workflows
– End-dates, then deletes notifications not referenced in
WF_ITEM_ACTIVITY_STATUSES, _H (orphaned)
• Example: notifications from finished concurrent programs
– Purges ad-hoc roles where ORIG_SYSTEM =
‘WF_LOCAL_ROLES’ or ‘WF_LOCAL_USERS’ and
not referenced in WF_ROLE_HIERARCHIES or
WF_NOTIFICATIONS or WF_ITEMS.OWNER_ROLE
From
Finished
Concurrent
Requests
Catch Up Purging
Be Careful!!
• Purge by Item Type or Age to reduce time of each run
– Commit keeps rollback from being an issue
• Run with “Core Workflow Only” = Y
• After catching up, run one more time with “Core
Workflow Only” = N to purge old design data and
orphaned notifications
– Then only need “N” once/week (orphan notifications)
• While 10g, 11g automatically reset high water marks,
removing empty space may still be recommended
Rebuild or not to Rebuild
Starting
Point
After “Catch-
After “Catch-
up” Purge
up” Purge
Export/Import
More Activity More Activity

Regular Regular
Purges Purges
More Activity More Activity

Regular Regular
Purges Purges
Purge Issues
Synchronous Persistence Type
• Synchronous is not a choice for Persistence Type
• Lookup code, Persistence Type, is Access Level
System (i.e. can’t add codes)
• Solution
– Help | Diagnostics | Examine
• Field: CUSTOMIZATION_LEVEL
• Value: E
– Add Lookup Code for Sync
– Restore CUSTOMIZATION_LEVEL to S
Purge Issues
Completed With Errors
• ‘Completed with Errors’ – end date is set in
WF_ITEMS but at least one activity exists in
WF_ACTIVITY_STATUSES with status of ERROR or
ACTIVE
– Can’t abort, can’t purge, and if linked to parent or
child, parent or child cannot be purged
Same workflow
Each Error has at
least 1 Active
Purge Issues
• Modified version of $FND_TOP/sql/wfrmtype.sql
Purge Issues
Purge – MOS Notes
• 337923.1 “A closer examination of the Concurrent

Program Purge Obsolete Workflow Runtime Data”
• 132254.1 “Speeding Up and Purging Workflows”
• 298550.1 “Troubleshooting Workflow Data Growth
Issues”
• 144806.1 “A Detailed Approach to Purging Oracle
Workflow Runtime Data”
• 277124.1 “FAQ on Purging Oracle Workflow Data”
Note: Most referenced patches already in 11i.10, R12
Purge – MOS Notes
Parent / Child Problems
• 752383.1 “Purge Obsolete Workflow Runtime Data
Concurrent Request (FNDWFPR) Is Not Purging Data”
SELECT c.item_type child,
DECODE (c.end_date, NULL,'OPEN','CLOSED') child_status,
c.parent_item_type parent,
DECODE (c.parent_item_type, NULL, 'NOPARENT',
DECODE (p.end_date,NULL,'OPEN',
'CLOSED'))parent_status,
COUNT (*)
FROM wf_items p, wf_items c
WHERE p.item_type(+) = c.parent_item_type Business
AND p.item_key(+) = c.parent_item_key
GROUP BY c.item_type, DECODE(c.end_date,NULL,'OPEN',CLOSED'), Event
c.parent_item_type, Errors
DECODE (c.parent_item_type, NULL, 'NOPARENT',
DECODE (p.end_date, NULL, 'OPEN', 'CLOSED'))
ORDER BY c.item_type, c.parent_item_type;
Purge – MOS Notes
Parent / Child Problems
• 1378954.1 “bde_wf_process_tree.sql – For analyzing
the Root, Children, Grandchildren Associations of a
Single Workflow”
Purge – MOS Notes
Product Specific
• Order Management
– 751026.1 “FNDWFPR – Purge Obsolete Workflow Runtime
Data – OEOH / OEOL Performance Issues”
• Scripts to help close attached children
– 398822.1 “Order Management Suite – Data Fix Script
Patch”
– 405275.1 “How to Detect Data Corruption and Purge More
Eligible OEOH/OEOL Workflow Items for Order
Management Workflow”
– 878032.1 “How To Use Concurrent Program “Purge Order
Management Workflow””
• 11i.10 – Patch 9845873
• 12.1.2+ – included (not available for 12.0.x)
• Purges closed lines even when header still open
• “Attempt to Close” – purges OMERROR / WFERROR that are
orphan or attached to no-longer-error nodes
Purge – MOS Notes
Product Specific
• Purchasing
– 458886.1 “How To Diagnose Issues Related To Purge
Of Purchasing Workflow Data That Remain Even After
Running The ‘Purge Obsolete Workflow Runtime Data’
Concurrent Program?”
• Scripts to help close attached children
Background Processing
Background Engines
• Run Engine for Stuck separately

– Parameters NULL, NULL, NULL, No, No, Yes
– Run once/week or once/month
• Run Engine for Timed Out activites separately based
on criticality of timeout
– If average timeout = 1 day, run once/day
– Parameters NULL, NULL, NULL, No, Yes, No
Background Engines
• Run Engine for Deferred activities separately based on

criticality of activity
– Except for OEOL, very few workflows need moving
more than every 15 minutes
– If Order volume high, run “targeted” engine for OEOL
every 5 minutes
• Parameters: Order Line, NULL, NULL, Yes, No, No
– Run generic every 15-60 minutes
• Parameters: NULL, NULL, NULL, Yes, No, No
‘APPS’
not part of
workflow
name
• Note counts of Ready, run SQL and see if queue is
steady or growing
– Growing – decrease wait time for next execution of
Background Engine
– Low or Empty – increase wait time
Background Engines
• Activities in queue table WF_DEFERRED_TABLE_M

– Time to process = DEQ_TIME – ENQ_TIME where
STATE=2
• By default, data only stays in this table 24 hours after
DEQ_TIME
• In 12.2, these fields are datatype TIMESTAMP
• 369537.1 “How to Monitor the FNDWFBG – Workflow
Background Program”
– Scripts: what’s in queue, what will be dequeued next
• 466535.1 “How to Resolve the Most Common
Workflow Background Engine Problems”
• Check MOS using “Background Process Performance”
for notes for specific item type issues
Background Engine Runs a
Long Time
• 186361.1 “WF 2.x: Workflow Background Process
Performance Troubleshooting Guide”
– Determine the Item Type Causing the Issue
• Usually it’s the activity being processed, not the
Background Engine
• Check for Loops in Workflow – see next slide
• 560144.1 “11.5.10.4: Workflow Background Process
Seems to Take Longer After Rup4”
– Don’t use re-submit time < 5 minutes
– AQ_TM_PROCESSES must be autotune or at least 1
• Never set > 9
Loops in Workflow
(from Workflow Analyzer)
AQ_TM_PROCESSES
Background Engine Runs a
Long Time
• 560144.1 “11.5.10.4: Workflow Background Process
Seems to Take Longer After Rup4” (cont)
– JOB_QUEUE_PROCESSES at least 5
• OAM recommends 10
• Oracle seeds this to 2 – change it (ignore note
396009.1, updated Feb 2012)
• MOS note 1530928.1, dated Feb 2013, value of 2
causes Row Lock Contention on Mailer
• MOS note 578831.1 – explains how to monitor
• MOS note 271855.1 – if database prior to 11.2,
perform regular rebuilds/coalesces on all dequeue
indexes/IOTs associated with an AQ table
JOB_QUEUE_PROCESSES
• 1196298.1 – Don’t explicitly set value to zero

Queue Performance
Advanced Queue Performance
• 469009.1 “Troubleshooting Workflow Agent Listener’s

failure to start”
• 741087.1 “High Logging Messages on
WF_EVEN_OJMSTEXT_QH procedure”
– Verify Profile options (issue is level 2, 3 messages)
• FND: Debug Log Level – Unexpected (level 6)
– Note: 1107970.1 recommends setting
» FND: Debug Log Enables – Yes
» FND: Debug Module = %
– Schedule “Purge Diagnostic and Log Messages”
• Set Log Level for each Listener to Error, then stop
and restart Workflow Agent Listener Container
• Memory insufficient or Containers consuming all

available memory
– 444939.1 “How do you Change the Maximum Memory
Size taken by Workflow Service Container”
• Retention
– Increases performance if = 0, but destroys ability to
tune, troubleshoot
– Recommend 1 day – 86400 seconds for the following
queues for troubleshooting
• WF_ERROR, WF_JAVA_ERROR, WF_DEFFERED,
WF_JAVA_DEFERRED, WF_NOTIFICATION_IN/OUT
– Dbms-aqadm.alter_queue(queue_name =
>’<queue>’,retention_time=>86400);
• White paper “Application Development with Oracle

Advanced Queuing” by Jeff Jacobs (works for Paypal -
lots of queue traffic)
– www.nocoug.org/download/2012-
08/Jeff_Jacobs_Advanced_Queueing.pdf
– The basic solution is:
• Add the index to CORRID column
• Get the CBO to recognize it, which can be by
generating an appropriate set of statistics or the
various forms of SQL plan management
WF_CONTROL
• Controls all other queues
• Run ‘Control Queue Cleanup’ every 12 hours
– In every instance
• 469045.1 “Troubleshooting WF_CONTROL Agent
Issues”
– Discussion of this queue
– Scripts to run to ensure subscribers are valid and dead
subscribers are removed properly
WF_DEFERRED
• Subscriptions to Events Phase > 100
• Workflows started by events
• 334348.1 “Low Performance Processing Messages in
WF_DEFERRED Queue”; 468650.1 “Troubleshooting
WF_DEFERRED Agent Listeners Performance”
– Use SQL to determine Events in queue
– Identify if events not being dequeued in timely fashion
• Time in queue>2X sleep time for queue
– Identify events with long processing time
• Trace code and identify issues (bugs, tuning, etc)
WF_DEFERRED
• 334348.1 “Low Performance Processing Message in
WF_DEFERRED Queue”; 467650.1 “Troubleshooting
WF_DEFERRED Agent Listeners Performance” (cont)
– Identify Events with high volume
• Create additional generic agent listeners
• Create specific agent listeners
• Increase ‘Inbound Thread Count’
(PROCESSOR_IN_THREAD_COUNT) by 1 until
performance acceptable
– Temporarily set retention time to 0
WF_DEFERRED
• Queue may be corrupt
– Receiving Errors “ORA-24033:No Recipients for
Message”
– Rebuild using instructions in note 286394.1 “How to
rebuild the WF_DEFERRED queue”
Mailer Performance
Notification Mailer
Click icon, change

Status to Disabled
• If global preference is ‘Do not send me mail’ (QUERY)

– Use Framework Personalization – prohibit override from
Preferences link
• Ensure records in FND_USER_PREFERENCES updated to
QUERY
– For event oracle.apps.wf.notification.send.group, disable local
subscription w/Out Agent WF_NOTIFICATION@
• Do this in test/dev or if you are not using the mailer at all
– 453137.1 “Oracle Workflow Best Practices Release 12 and
Release 11i”
• Remember!! Alert now uses the Mailer
Notification Mailer
• If global preference is ‘Do Not Send Me Mail’ and not

running Alert
– Don’t start Mailer
– Set Startup mode for following listeners to Manual or
On Demand
• Workflow Deferred Notification Agent Listener
• Workflow Inbound Notifications Agent Listener
• Monitor WF_NOTIFICATION_IN, _OUT
• Monitor WF_DEFERRED for
oracle.apps.wf.notification.% events
Notification Mailer
• If Inbound Processing is not checked and not running

Alert inbound processing
– Set Startup mode for following listeners to Manual or
On Demand
• Workflow Inbound Notifications Agent Listener
• Monitor WF_NOTIFICATION_IN
• Mailer only for Alert
– 463777.1 “How to Disable all Workflow related Email
Notifications Except for the Ones Sent from Oracle
Alerts?”
– Create new Mailer
• Set Correlation id = ALR:%
Notification Mailer
• Increase Inbound Polling Interval – Processor Min

Loop Sleep (seconds) – ensure processor Max Loop
Sleep at least 5*Processor Min Loop Sleep
– Note 315748.1 “How to Change The Java Workflow
Mailer Inbound Polling Interval”
Notification Mailer
Click it,
issues
outweigh
benefits
• Processor Close on Read Timeout
– 315748.1 – unclick for performance
• We don’t agree with this note
– 422870.1 – unless clicked, not removed from Process
folder
– 332152.1 – must be clicked if running multiple mailers
using same SMTP Server (Outbound Name) or will
get contention and locking
– 437986.1 – must be clicked or messages get stuck in
Inbox
Notification Mailer
• Mailer Log shows java.lang.OutOfMemoryError

– 467516.1 “Users suddently (sic) Stop Receiving Email
Notifications”
• Insufficient Heap Size (Xmx and Xms)
• Edit $APPL_TOP/admin/adovars.env
– Add/change following
» APPSJREOPT=“-Xms 128m-Xmx3072m”
» Export APPSJREOPT
• Bounce Concurrent Managers
Notification Mailer
• You Have Insufficient Privileges”

– 414376.1 “”You Have Insufficient Privileges For The
Current Operation” On Reqapprv Notif”
– Create dedicated user for the mailer
• Framework URL
timeout = 120
0 is SYSADMIN
Notification Mailer
• Setup separate user to run the Mailer

– Must be a workflow administrator
• This forces Administrator to be a responsibility as
SYSADMIN must ALWAYS be an administrator
– Should only have following responsibilities
• System Administrator
• Responsibility used as workflow administrator
– Should not be a user with other duties
• Why not SYSADMIN
– Performance: SYSADMIN usually has too many of own
emails due to WFERROR emails
– Manageability: Enabling log for SYSADMIN includes many
other functions than mailer thus hampering troubleshooting
Notification Mailer
• Tag Files
– Out of Office, Undeliverable – set to Ignore
• 388709.1 “Email Notification Failures Are Causing
The Email Servers To Crash”
• Uncheck Mailer parameter “Send warning for
unsolicited e-mail”
– 431359.1 “Setting up a Tag in the Mailer configuration
files to handle unsolicited mail”
• Uncheck Mailer parameter “Send e-mails for canceled
notifications”
Miscellaneous
Performance Aides
Item Attributes “As Needed”
• By default, when workflow initiated, runtime copy of each

item attribute created
• 66% if item attributes have no value (and that excludes
Event attributes)
SELECT COUNT (*)
,v.item_type
FROM wf_item_attribute_values v
,wf_item_attributes a
WHERE a.item_type = v.item_type
AND a.NAME = v.NAME
AND a.TYPE <> 'EVENT'
AND v.text_value IS NULL
AND v.number_value IS NULL
AND v.date_value IS NULL
GROUP BY v.item_type
ORDER BY 1 DESC;
Item Attributes “As Needed”
• #ONDEMANDATTR
– Process Activity Attribute
– For top-level runnable process
– Type must be Text or Lookup
• Use Yes/No Lookup Type
• For Text, Value must be Y
– Do not assign an item attribute as the value
– Runtime copy only created when SetItemAttr<> used
• If referenced prior to this call, default value used
– Experiment with a particular workflow
• HRSSA, REQAPPRV, POAPPRV, etc
Profile Options
• Account Generator:Run in Debug Mode

– Except when experiencing an issue with Account
Generator, set it to ‘No’
– Make sure when problem fixed to purge workflows and
reset (TEMP and PERM)
• PO:Workflow Processing Mode
– If set to ‘Online’, screen does not return control to
Buyer until workflow ends or notifications requiring
response is encountered
• Still requires background engine to complete
– If Buyers cannot self-approve POs, set to ‘Background’
Profile Options – R12.2
• WF: Defer Processing of Notification Response

– Behavior when responding to notifications through
worklist
• No– control not returned to screen until workflow
ended or reached next deferred activity
• Yes – control returned immediately, workflow
continues when background engine runs again
• Specific – behaves as Yes for item types specified in
Lookup Type “Workflow Notification Response Deferred
Item Types”
– Can also be set from
Workflow Administration
page
Workflow Statistics Programs
• Don’t set these to run every 5 minutes

– Workflow Mailer Statistics Concurrent Program
– Workflow Work Items Statistics Concurrent Program
– Workflow Agent Activity Statistics Concurrent Program
• Run Once/Day
– Just in case you forget to click icon to refresh queries
• Run before running Workflow Analyzer – analyzer
uses data from these programs
• 787228.1 “Cannot Abort Old Open Items in Workflow
Manager Because Errored Items are not Returned”
– 12.0.4 – wf_item_types.num_error=0, won’t show
– 12.0.6 – click refresh button and is re-calculated
Workflow Concurrent
Managers
• Workflow Agent Listener Service (WFALSNRSVC)
must be enabled and active – ALWAYS!!
• Workflow Mailer Service (WFMLRSVC) must be
enabled if emailing notifications or running Alert
• Workflow Document Web Services Service
(WFWSSVC) must be enabled to use Web Services
Pinning
• Objects “pinned” into memory so they do not need to

be constantly reloaded from disk, flushed out of
memory and reloaded
– PIND
– 301171.1 “Toolkit for dynamic marking of Library
Cache objects as Kept (PIND)”
– Requires large SGA and memory
• Consider pinning WF_ packages
– There are approximately 120 packages, not all used
by every organization
– Start with WF_ENGINE
Run 64-bit Database
• Memory is critical, 32-bit can’t address enough

• If not on 11gR2, Upgrade
– Always monitor database desupport policies
• Different from EBS desupport policies
Wffngen.sql
• Translates activity function calls into static calls

– According to Oracle, 25% increase in performance
• Look for variable itemtypeList_t
– Seeded := itemtypeList_t (‘WFSTD’,’FNDFFWF’)
– Add following item types (after configuration complete)
• WFERROR, POERROR, OMERROR
• Other workflows with high (current) count in
WF_ITEMS
• Reference
– How to Improve Performance in Workflow? (Doc ID
1524731.1)
Order Management
• 130511.1 “PERFORMANCE Issues in OM, SE, QP”

– Remove unnecessary activities
– No sub-processes
– Make Scheduling a deferred activity
• Consider seeded Line process: “Line Flow Generic:
Performance”
– Removes unnecessary activities and sub-processes,
reducing WF data significantly
Partition Tables
• Perform after Purge cleanup

– Doing this replaces need to export/import
• Backup following tables
– WF_ITEM_ACTIVITY_STATUSES
– WF_ITEM_ACTIVITY_STATUSES_H
– WF_ITEM_ATTRIBUTE_VALUES
– WF_ITEMS
• Ensure have free space in same tablespace slightly
more than currently used (incl. indices)
• Move to OATM first – 402720.1 “OATM Migration fails
with ORA-14257 when moving list partitioned tables”
Partition Tables
• If using RAC, set up RAC Affinity

– After partitioning
• Create VPD policy
• Configure workflow processes
• Schedule “Workflow Background Process for RAC”
• If not using RAC, partition using WFPART.sql
– Apply patch 7252442 (11i – note 749105.1) or patch
8241676 (R12.1 – note 789528.1) afterwards to
create missing indices
• See R12.2 Workflow Administrator’s Guide, Chapter 2
for detailed steps for either method
– This guide can be used in 11i and R12.1 for non-RAC
partitioning instructions
Workflow Book
Available from
Amazon.com,
Barnes & Noble
(bn.com)
Lulu.com
(pdf version)
Thank you!!
Karen.Brownfield@Infosemantics.com

Workflow Performance Tuning in Release 12.1

Uploaded by

Document Informationclick to expand document informationOracle WF performance tuning in R12.1

Document Informationclick to expand document information

Copyright:

Available Formats

Workflow Performance Tuning in Release 12.1

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Workflow Performance Tuning in Release 12.1

Uploaded by

Copyright:

Available Formats

Workflow Performance Tuning

• 1369938.1 “Workflow Analyzer script for E-Business

• Sign up for email notifications

• It’s not just the RUPs, one-offs are important

Despite low overall

• Perform following query

• Play with Order By – try 5, 1, 2 also

• Triage – Most Recent, Highest Numbers

• Purge now closes WFERROR where calling activity is

• Perform following query

• Find and fix what causes event to error

• Message to SYSADMIN can re-raise event if still needs

• Adjust $FND_TOP/sql/wfrmtype.sql script

Click the link to the sql

HR added v4.09, PO v5.01

11i Purgeable for

• Need schedule for Temporary and for Permanent

• Aborts WFERROR where PARENT_ITEM_TYPE

• (R12.2) Other Cached Data = Y

• If “Core Workflow Only” = N

More Activity More Activity

More Activity More Activity

• 337923.1 “A closer examination of the Concurrent

Note: Most referenced patches already in 11i.10, R12

• Run Engine for Stuck separately

• Run Engine for Deferred activities separately based on

• Activities in queue table WF_DEFERRED_TABLE_M

• 1196298.1 – Don’t explicitly set value to zero

• 469009.1 “Troubleshooting Workflow Agent Listener’s

• Memory insufficient or Containers consuming all

• White paper “Application Development with Oracle

Click icon, change

• If global preference is ‘Do not send me mail’ (QUERY)

• If global preference is ‘Do Not Send Me Mail’ and not

• If Inbound Processing is not checked and not running

• Increase Inbound Polling Interval – Processor Min

• Mailer Log shows java.lang.OutOfMemoryError

• You Have Insufficient Privileges”

• Setup separate user to run the Mailer

• By default, when workflow initiated, runtime copy of each

• Account Generator:Run in Debug Mode

• WF: Defer Processing of Notification Response

• Don’t set these to run every 5 minutes

• Objects “pinned” into memory so they do not need to

• Memory is critical, 32-bit can’t address enough

• Translates activity function calls into static calls

• 130511.1 “PERFORMANCE Issues in OM, SE, QP”

• Perform after Purge cleanup

• If using RAC, set up RAC Affinity

You might also like