Enable parallel SELECT for "INSERT INTO ... SELECT ...".

Parallel SELECT can't be utilized for INSERT in the following cases:
- INSERT statement uses the ON CONFLICT DO UPDATE clause
- Target table has a parallel-unsafe: trigger, index expression or
  predicate, column default expression or check constraint
- Target table has a parallel-unsafe domain constraint on any column
- Target table is a partitioned table with a parallel-unsafe partition key
  expression or support function

The planner is updated to perform additional parallel-safety checks for
the cases listed above, for determining whether it is safe to run INSERT
in parallel-mode with an underlying parallel SELECT. The planner will
consider using parallel SELECT for "INSERT INTO ... SELECT ...", provided
nothing unsafe is found from the additional parallel-safety checks, or
from the existing parallel-safety checks for SELECT.

While checking parallel-safety, we need to check it for all the partitions
on the table which can be costly especially when we decide not to use a
parallel plan. So, in a separate patch, we will introduce a GUC and or a
reloption to enable/disable parallelism for Insert statements.

Prior to entering parallel-mode for the execution of INSERT with parallel
SELECT, a TransactionId is acquired and assigned to the current
transaction state. This is necessary to prevent the INSERT from attempting
to assign the TransactionId whilst in parallel-mode, which is not allowed.
This approach has a disadvantage in that if the underlying SELECT does not
return any rows, then the TransactionId is not used, however that
shouldn't happen in practice in many cases.

Author: Greg Nancarrow, Amit Langote, Amit Kapila
Reviewed-by: Amit Langote, Hou Zhijie, Takayuki Tsunakawa, Antonin Houska, Bharath Rupireddy, Dilip Kumar, Vignesh C, Zhihong Yu, Amit Kapila
Tested-by: Tang, Haiying
Discussion: https://postgr.es/m/CAJcOf-cXnB5cnMKqWEp2E2z7Mvcd04iLVmV=qpFJrR3AcrTS3g@mail.gmail.com
Discussion: https://postgr.es/m/CAJcOf-fAdj=nDKMsRhQzndm-O13NY4dL6xGcEvdX5Xvbbi0V7g@mail.gmail.com

Author: Amit Kapila

doc/src/sgml/parallel.sgml

@@ -146,7 +146,9 @@ EXPLAIN SELECT * FROM pgbench_accounts WHERE filler LIKE '%x%';
         a CTE, no parallel plans for that query will be generated.  As an
         exception, the commands <literal>CREATE TABLE ... AS</literal>, <literal>SELECT
         INTO</literal>, and <literal>CREATE MATERIALIZED VIEW</literal> which create a new
-        table and populate it can use a parallel plan.
+        table and populate it can use a parallel plan. Another exception is the command
+        <literal>INSERT INTO ... SELECT ...</literal> which can use a parallel plan for
+        the underlying <literal>SELECT</literal> part of the query.
       </para>
     </listitem>
 

src/backend/access/transam/xact.c

@@ -1014,6 +1014,32 @@ IsInParallelMode(void)
 	return CurrentTransactionState->parallelModeLevel != 0;
 }
 
+/*
+ *	PrepareParallelModePlanExec
+ *
+ * Prepare for entering parallel mode plan execution, based on command-type.
+ */
+void
+PrepareParallelModePlanExec(CmdType commandType)
+{
+	if (IsModifySupportedInParallelMode(commandType))
+	{
+		Assert(!IsInParallelMode());
+
+		/*
+		 * Prepare for entering parallel mode by assigning a TransactionId.
+		 * Failure to do this now would result in heap_insert() subsequently
+		 * attempting to assign a TransactionId whilst in parallel-mode, which
+		 * is not allowed.
+		 *
+		 * This approach has a disadvantage in that if the underlying SELECT
+		 * does not return any rows, then the TransactionId is not used,
+		 * however that shouldn't happen in practice in many cases.
+		 */
+		(void) GetCurrentTransactionId();
+	}
+}
+
 /*
  *	CommandCounterIncrement
  */

src/backend/executor/execMain.c

@@ -1512,7 +1512,10 @@ ExecutePlan(EState *estate,
 
 	estate->es_use_parallel_mode = use_parallel_mode;
 	if (use_parallel_mode)
+	{
+		PrepareParallelModePlanExec(estate->es_plannedstmt->commandType);
 		EnterParallelMode();
+	}
 
 	/*
 	 * Loop until we've processed the proper number of tuples from the plan.

src/backend/nodes/copyfuncs.c

@@ -96,6 +96,7 @@ _copyPlannedStmt(const PlannedStmt *from)
 	COPY_BITMAPSET_FIELD(rewindPlanIDs);
 	COPY_NODE_FIELD(rowMarks);
 	COPY_NODE_FIELD(relationOids);
+	COPY_NODE_FIELD(partitionOids);
 	COPY_NODE_FIELD(invalItems);
 	COPY_NODE_FIELD(paramExecTypes);
 	COPY_NODE_FIELD(utilityStmt);

src/backend/nodes/outfuncs.c

@@ -314,6 +314,7 @@ _outPlannedStmt(StringInfo str, const PlannedStmt *node)
 	WRITE_BITMAPSET_FIELD(rewindPlanIDs);
 	WRITE_NODE_FIELD(rowMarks);
 	WRITE_NODE_FIELD(relationOids);
+	WRITE_NODE_FIELD(partitionOids);
 	WRITE_NODE_FIELD(invalItems);
 	WRITE_NODE_FIELD(paramExecTypes);
 	WRITE_NODE_FIELD(utilityStmt);
@@ -2221,6 +2222,7 @@ _outPlannerGlobal(StringInfo str, const PlannerGlobal *node)
 	WRITE_NODE_FIELD(resultRelations);
 	WRITE_NODE_FIELD(appendRelations);
 	WRITE_NODE_FIELD(relationOids);
+	WRITE_NODE_FIELD(partitionOids);
 	WRITE_NODE_FIELD(invalItems);
 	WRITE_NODE_FIELD(paramExecTypes);
 	WRITE_UINT_FIELD(lastPHId);

src/backend/nodes/readfuncs.c

@@ -1590,6 +1590,7 @@ _readPlannedStmt(void)
 	READ_BITMAPSET_FIELD(rewindPlanIDs);
 	READ_NODE_FIELD(rowMarks);
 	READ_NODE_FIELD(relationOids);
+	READ_NODE_FIELD(partitionOids);
 	READ_NODE_FIELD(invalItems);
 	READ_NODE_FIELD(paramExecTypes);
 	READ_NODE_FIELD(utilityStmt);

src/backend/optimizer/plan/planner.c

@@ -305,6 +305,7 @@ standard_planner(Query *parse, const char *query_string, int cursorOptions,
 	glob->resultRelations = NIL;
 	glob->appendRelations = NIL;
 	glob->relationOids = NIL;
+	glob->partitionOids = NIL;
 	glob->invalItems = NIL;
 	glob->paramExecTypes = NIL;
 	glob->lastPHId = 0;
@@ -316,16 +317,16 @@ standard_planner(Query *parse, const char *query_string, int cursorOptions,
 	/*
 	 * Assess whether it's feasible to use parallel mode for this query. We
 	 * can't do this in a standalone backend, or if the command will try to
-	 * modify any data, or if this is a cursor operation, or if GUCs are set
-	 * to values that don't permit parallelism, or if parallel-unsafe
-	 * functions are present in the query tree.
+	 * modify any data (except for Insert), or if this is a cursor operation,
+	 * or if GUCs are set to values that don't permit parallelism, or if
+	 * parallel-unsafe functions are present in the query tree.
 	 *
-	 * (Note that we do allow CREATE TABLE AS, SELECT INTO, and CREATE
-	 * MATERIALIZED VIEW to use parallel plans, but as of now, only the leader
-	 * backend writes into a completely new table.  In the future, we can
-	 * extend it to allow workers to write into the table.  However, to allow
-	 * parallel updates and deletes, we have to solve other problems,
-	 * especially around combo CIDs.)
+	 * (Note that we do allow CREATE TABLE AS, INSERT INTO...SELECT, SELECT
+	 * INTO, and CREATE MATERIALIZED VIEW to use parallel plans. However, as
+	 * of now, only the leader backend writes into a completely new table. In
+	 * the future, we can extend it to allow workers to write into the table.
+	 * However, to allow parallel updates and deletes, we have to solve other
+	 * problems, especially around combo CIDs.)
 	 *
 	 * For now, we don't try to use parallel mode if we're running inside a
 	 * parallel worker.  We might eventually be able to relax this
@@ -334,13 +335,14 @@ standard_planner(Query *parse, const char *query_string, int cursorOptions,
 	 */
 	if ((cursorOptions & CURSOR_OPT_PARALLEL_OK) != 0 &&
 		IsUnderPostmaster &&
-		parse->commandType == CMD_SELECT &&
+		(parse->commandType == CMD_SELECT ||
+		 is_parallel_allowed_for_modify(parse)) &&
 		!parse->hasModifyingCTE &&
 		max_parallel_workers_per_gather > 0 &&
 		!IsParallelWorker())
 	{
 		/* all the cheap tests pass, so scan the query tree */
-		glob->maxParallelHazard = max_parallel_hazard(parse);
+		glob->maxParallelHazard = max_parallel_hazard(parse, glob);
 		glob->parallelModeOK = (glob->maxParallelHazard != PROPARALLEL_UNSAFE);
 	}
 	else
@@ -521,6 +523,19 @@ standard_planner(Query *parse, const char *query_string, int cursorOptions,
 	result->rewindPlanIDs = glob->rewindPlanIDs;
 	result->rowMarks = glob->finalrowmarks;
 	result->relationOids = glob->relationOids;
+
+	/*
+	 * Register the Oids of parallel-safety-checked partitions as plan
+	 * dependencies. This is only really needed in the case of a parallel plan
+	 * so that if parallel-unsafe properties are subsequently defined on the
+	 * partitions, the cached parallel plan will be invalidated, and a
+	 * non-parallel plan will be generated.
+	 *
+	 * We also use this list to acquire locks on partitions before executing
+	 * cached plan. See AcquireExecutorLocks().
+	 */
+	if (glob->partitionOids != NIL && glob->parallelModeNeeded)
+		result->partitionOids = glob->partitionOids;
 	result->invalItems = glob->invalItems;
 	result->paramExecTypes = glob->paramExecTypes;
 	/* utilityStmt should be null, but we might as well copy it */