From 26ae52fb9e9ecb4475b7db71773da92ea86ad8a4 Mon Sep 17 00:00:00 2001 From: Lukas Fittl Date: Tue, 9 Sep 2025 02:16:59 -0700 Subject: [PATCH v11 5/9] Optimize measuring WAL/buffer usage through stack-based instrumentation Previously, in order to determine the buffer/WAL usage of a given code section, we utilized continuously incrementing global counters that get updated when the actual activity (e.g. shared block read) occurred, and then calculated a diff when the code section ended. This resulted in a bottleneck for executor node instrumentation specifically, with the function BufferUsageAccumDiff showing up in profiles and in some cases adding up to 10% overhead to an EXPLAIN (ANALYZE, BUFFERS) run. Instead, introduce a stack-based mechanism, where the actual activity writes into the current stack entry. In the case of executor nodes, this means that each node gets its own stack entry that is pushed at InstrStartNode, and popped at InstrEndNode. Stack entries are zero initialized (avoiding the diff mechanism) and get added to their parent entry when they are finalized, i.e. no more modifications can occur. To correctly handle abort situations, any use of instrumentation stacks must involve either a top-level QueryInstrumentation struct, and its associated InstrQueryStart/InstrQueryStop helpers (which use resource owners to handle aborts), or the Instrumentation struct itself with dedicated PG_TRY/PG_FINALLY calls that ensure the stack is in a consistent state after an abort. This also drops the global pgBufferUsage, any callers interested in measuring buffer activity should instead utilize InstrStart/InstrStop. The related global pgWalUsage is kept for now due to its use in pgstat to track aggregate WAL activity and heap_page_prune_and_freeze for measuring FPIs. Author: Lukas Fittl Reviewed-by: Zsolt Parragi Reviewed-by: Heikki Linnakangas Discussion: https://www.postgresql.org/message-id/flat/CAP53PkxrmpECzVFpeeEEHDGe6u625s%2BYkmVv5-gw3L_NDSfbiA%40mail.gmail.com#cb583a08e8e096aa1f093bb178906173 --- contrib/auto_explain/auto_explain.c | 16 +- .../pg_stat_statements/pg_stat_statements.c | 87 +--- src/backend/access/brin/brin.c | 10 +- src/backend/access/gin/gininsert.c | 10 +- src/backend/access/heap/vacuumlazy.c | 15 +- src/backend/access/nbtree/nbtsort.c | 10 +- src/backend/commands/analyze.c | 31 +- src/backend/commands/explain.c | 43 +- src/backend/commands/explain_dr.c | 57 ++- src/backend/commands/prepare.c | 27 +- src/backend/commands/tablecmds.c | 2 +- src/backend/commands/trigger.c | 17 +- src/backend/commands/vacuumparallel.c | 10 +- src/backend/executor/README.instrument | 227 +++++++++ src/backend/executor/execMain.c | 83 +++- src/backend/executor/execParallel.c | 32 +- src/backend/executor/execPartition.c | 2 +- src/backend/executor/execProcnode.c | 84 +++- src/backend/executor/execUtils.c | 11 +- src/backend/executor/instrument.c | 448 +++++++++++++----- src/backend/replication/logical/worker.c | 2 +- src/backend/storage/buffer/bufmgr.c | 6 +- src/backend/utils/activity/pgstat_io.c | 6 +- src/include/commands/explain_dr.h | 5 +- src/include/executor/execdesc.h | 4 +- src/include/executor/executor.h | 5 +- src/include/executor/instrument.h | 198 +++++++- src/include/nodes/execnodes.h | 3 +- src/include/utils/resowner.h | 1 + src/tools/pgindent/typedefs.list | 2 + 30 files changed, 1097 insertions(+), 357 deletions(-) create mode 100644 src/backend/executor/README.instrument diff --git a/contrib/auto_explain/auto_explain.c b/contrib/auto_explain/auto_explain.c index 39bf2543b70..4be81489ff4 100644 --- a/contrib/auto_explain/auto_explain.c +++ b/contrib/auto_explain/auto_explain.c @@ -305,19 +305,9 @@ explain_ExecutorStart(QueryDesc *queryDesc, int eflags) if (auto_explain_enabled()) { - /* - * Set up to track total elapsed time in ExecutorRun. Make sure the - * space is allocated in the per-query context so it will go away at - * ExecutorEnd. - */ + /* Set up to track total elapsed time in ExecutorRun. */ if (queryDesc->totaltime == NULL) - { - MemoryContext oldcxt; - - oldcxt = MemoryContextSwitchTo(queryDesc->estate->es_query_cxt); - queryDesc->totaltime = InstrAlloc(INSTRUMENT_ALL); - MemoryContextSwitchTo(oldcxt); - } + queryDesc->totaltime = InstrQueryAlloc(INSTRUMENT_ALL); } } @@ -382,7 +372,7 @@ explain_ExecutorEnd(QueryDesc *queryDesc) oldcxt = MemoryContextSwitchTo(queryDesc->estate->es_query_cxt); /* Log plan if duration is exceeded. */ - msec = INSTR_TIME_GET_MILLISEC(queryDesc->totaltime->total); + msec = INSTR_TIME_GET_MILLISEC(queryDesc->totaltime->instr.total); if (msec >= auto_explain_log_min_duration) { ExplainState *es = NewExplainState(); diff --git a/contrib/pg_stat_statements/pg_stat_statements.c b/contrib/pg_stat_statements/pg_stat_statements.c index 3e79108846e..9856dec3a5f 100644 --- a/contrib/pg_stat_statements/pg_stat_statements.c +++ b/contrib/pg_stat_statements/pg_stat_statements.c @@ -910,22 +910,11 @@ pgss_planner(Query *parse, && pgss_track_planning && query_string && parse->queryId != INT64CONST(0)) { - instr_time start; - instr_time duration; - BufferUsage bufusage_start, - bufusage; - WalUsage walusage_start, - walusage; + Instrumentation instr = {0}; - /* We need to track buffer usage as the planner can access them. */ - bufusage_start = pgBufferUsage; - - /* - * Similarly the planner could write some WAL records in some cases - * (e.g. setting a hint bit with those being WAL-logged) - */ - walusage_start = pgWalUsage; - INSTR_TIME_SET_CURRENT(start); + /* Track time and buffer/WAL usage as the planner can access them. */ + InstrInitOptions(&instr, INSTRUMENT_ALL); + InstrStart(&instr); nesting_level++; PG_TRY(); @@ -939,30 +928,20 @@ pgss_planner(Query *parse, } PG_FINALLY(); { + InstrStopFinalize(&instr); nesting_level--; } PG_END_TRY(); - INSTR_TIME_SET_CURRENT(duration); - INSTR_TIME_SUBTRACT(duration, start); - - /* calc differences of buffer counters. */ - memset(&bufusage, 0, sizeof(BufferUsage)); - BufferUsageAccumDiff(&bufusage, &pgBufferUsage, &bufusage_start); - - /* calc differences of WAL counters. */ - memset(&walusage, 0, sizeof(WalUsage)); - WalUsageAccumDiff(&walusage, &pgWalUsage, &walusage_start); - pgss_store(query_string, parse->queryId, parse->stmt_location, parse->stmt_len, PGSS_PLAN, - INSTR_TIME_GET_MILLISEC(duration), + INSTR_TIME_GET_MILLISEC(instr.total), 0, - &bufusage, - &walusage, + &instr.bufusage, + &instr.walusage, NULL, NULL, 0, @@ -1014,19 +993,9 @@ pgss_ExecutorStart(QueryDesc *queryDesc, int eflags) */ if (pgss_enabled(nesting_level) && queryDesc->plannedstmt->queryId != INT64CONST(0)) { - /* - * Set up to track total elapsed time in ExecutorRun. Make sure the - * space is allocated in the per-query context so it will go away at - * ExecutorEnd. - */ + /* Set up to track total elapsed time in ExecutorRun. */ if (queryDesc->totaltime == NULL) - { - MemoryContext oldcxt; - - oldcxt = MemoryContextSwitchTo(queryDesc->estate->es_query_cxt); - queryDesc->totaltime = InstrAlloc(INSTRUMENT_ALL); - MemoryContextSwitchTo(oldcxt); - } + queryDesc->totaltime = InstrQueryAlloc(INSTRUMENT_ALL); } } @@ -1088,10 +1057,10 @@ pgss_ExecutorEnd(QueryDesc *queryDesc) queryDesc->plannedstmt->stmt_location, queryDesc->plannedstmt->stmt_len, PGSS_EXEC, - INSTR_TIME_GET_MILLISEC(queryDesc->totaltime->total), + INSTR_TIME_GET_MILLISEC(queryDesc->totaltime->instr.total), queryDesc->estate->es_total_processed, - &queryDesc->totaltime->bufusage, - &queryDesc->totaltime->walusage, + &queryDesc->totaltime->instr.bufusage, + &queryDesc->totaltime->instr.walusage, queryDesc->estate->es_jit ? &queryDesc->estate->es_jit->instr : NULL, NULL, queryDesc->estate->es_parallel_workers_to_launch, @@ -1155,17 +1124,11 @@ pgss_ProcessUtility(PlannedStmt *pstmt, const char *queryString, !IsA(parsetree, ExecuteStmt) && !IsA(parsetree, PrepareStmt)) { - instr_time start; - instr_time duration; uint64 rows; - BufferUsage bufusage_start, - bufusage; - WalUsage walusage_start, - walusage; + Instrumentation instr = {0}; - bufusage_start = pgBufferUsage; - walusage_start = pgWalUsage; - INSTR_TIME_SET_CURRENT(start); + InstrInitOptions(&instr, INSTRUMENT_ALL); + InstrStart(&instr); nesting_level++; PG_TRY(); @@ -1181,6 +1144,7 @@ pgss_ProcessUtility(PlannedStmt *pstmt, const char *queryString, } PG_FINALLY(); { + InstrStopFinalize(&instr); nesting_level--; } PG_END_TRY(); @@ -1195,9 +1159,6 @@ pgss_ProcessUtility(PlannedStmt *pstmt, const char *queryString, * former value, which'd otherwise be a good idea. */ - INSTR_TIME_SET_CURRENT(duration); - INSTR_TIME_SUBTRACT(duration, start); - /* * Track the total number of rows retrieved or affected by the utility * statements of COPY, FETCH, CREATE TABLE AS, CREATE MATERIALIZED @@ -1209,23 +1170,15 @@ pgss_ProcessUtility(PlannedStmt *pstmt, const char *queryString, qc->commandTag == CMDTAG_REFRESH_MATERIALIZED_VIEW)) ? qc->nprocessed : 0; - /* calc differences of buffer counters. */ - memset(&bufusage, 0, sizeof(BufferUsage)); - BufferUsageAccumDiff(&bufusage, &pgBufferUsage, &bufusage_start); - - /* calc differences of WAL counters. */ - memset(&walusage, 0, sizeof(WalUsage)); - WalUsageAccumDiff(&walusage, &pgWalUsage, &walusage_start); - pgss_store(queryString, saved_queryId, saved_stmt_location, saved_stmt_len, PGSS_EXEC, - INSTR_TIME_GET_MILLISEC(duration), + INSTR_TIME_GET_MILLISEC(instr.total), rows, - &bufusage, - &walusage, + &instr.bufusage, + &instr.walusage, NULL, NULL, 0, diff --git a/src/backend/access/brin/brin.c b/src/backend/access/brin/brin.c index 2a0f8c8e3b8..1ceb2306954 100644 --- a/src/backend/access/brin/brin.c +++ b/src/backend/access/brin/brin.c @@ -2434,8 +2434,8 @@ _brin_begin_parallel(BrinBuildState *buildstate, Relation heap, Relation index, * and PARALLEL_KEY_BUFFER_USAGE. * * If there are no extensions loaded that care, we could skip this. We - * have no way of knowing whether anyone's looking at pgWalUsage or - * pgBufferUsage, so do it unconditionally. + * have no way of knowing whether anyone's looking at instrumentation, so + * do it unconditionally. */ shm_toc_estimate_chunk(&pcxt->estimator, mul_size(sizeof(WalUsage), pcxt->nworkers)); @@ -2886,6 +2886,7 @@ _brin_parallel_build_main(dsm_segment *seg, shm_toc *toc) Relation indexRel; LOCKMODE heapLockmode; LOCKMODE indexLockmode; + QueryInstrumentation *instr; WalUsage *walusage; BufferUsage *bufferusage; int sortmem; @@ -2935,7 +2936,7 @@ _brin_parallel_build_main(dsm_segment *seg, shm_toc *toc) tuplesort_attach_shared(sharedsort, seg); /* Prepare to track buffer usage during parallel execution */ - InstrStartParallelQuery(); + instr = InstrStartParallelQuery(); /* * Might as well use reliable figure when doling out maintenance_work_mem @@ -2950,7 +2951,8 @@ _brin_parallel_build_main(dsm_segment *seg, shm_toc *toc) /* Report WAL/buffer usage during parallel execution */ bufferusage = shm_toc_lookup(toc, PARALLEL_KEY_BUFFER_USAGE, false); walusage = shm_toc_lookup(toc, PARALLEL_KEY_WAL_USAGE, false); - InstrEndParallelQuery(&bufferusage[ParallelWorkerNumber], + InstrEndParallelQuery(instr, + &bufferusage[ParallelWorkerNumber], &walusage[ParallelWorkerNumber]); index_close(indexRel, indexLockmode); diff --git a/src/backend/access/gin/gininsert.c b/src/backend/access/gin/gininsert.c index e54782d9dd8..04cd53916ca 100644 --- a/src/backend/access/gin/gininsert.c +++ b/src/backend/access/gin/gininsert.c @@ -991,8 +991,8 @@ _gin_begin_parallel(GinBuildState *buildstate, Relation heap, Relation index, * and PARALLEL_KEY_BUFFER_USAGE. * * If there are no extensions loaded that care, we could skip this. We - * have no way of knowing whether anyone's looking at pgWalUsage or - * pgBufferUsage, so do it unconditionally. + * have no way of knowing whether anyone's looking at instrumentation, so + * do it unconditionally. */ shm_toc_estimate_chunk(&pcxt->estimator, mul_size(sizeof(WalUsage), pcxt->nworkers)); @@ -2117,6 +2117,7 @@ _gin_parallel_build_main(dsm_segment *seg, shm_toc *toc) Relation indexRel; LOCKMODE heapLockmode; LOCKMODE indexLockmode; + QueryInstrumentation *instr; WalUsage *walusage; BufferUsage *bufferusage; int sortmem; @@ -2185,7 +2186,7 @@ _gin_parallel_build_main(dsm_segment *seg, shm_toc *toc) tuplesort_attach_shared(sharedsort, seg); /* Prepare to track buffer usage during parallel execution */ - InstrStartParallelQuery(); + instr = InstrStartParallelQuery(); /* * Might as well use reliable figure when doling out maintenance_work_mem @@ -2200,7 +2201,8 @@ _gin_parallel_build_main(dsm_segment *seg, shm_toc *toc) /* Report WAL/buffer usage during parallel execution */ bufferusage = shm_toc_lookup(toc, PARALLEL_KEY_BUFFER_USAGE, false); walusage = shm_toc_lookup(toc, PARALLEL_KEY_WAL_USAGE, false); - InstrEndParallelQuery(&bufferusage[ParallelWorkerNumber], + InstrEndParallelQuery(instr, + &bufferusage[ParallelWorkerNumber], &walusage[ParallelWorkerNumber]); index_close(indexRel, indexLockmode); diff --git a/src/backend/access/heap/vacuumlazy.c b/src/backend/access/heap/vacuumlazy.c index f698c2d899b..c95e6801ead 100644 --- a/src/backend/access/heap/vacuumlazy.c +++ b/src/backend/access/heap/vacuumlazy.c @@ -637,8 +637,7 @@ heap_vacuum_rel(Relation rel, const VacuumParams params, TimestampTz starttime = 0; PgStat_Counter startreadtime = 0, startwritetime = 0; - WalUsage startwalusage = pgWalUsage; - BufferUsage startbufferusage = pgBufferUsage; + QueryInstrumentation *instr = NULL; ErrorContextCallback errcallback; char **indnames = NULL; Size dead_items_max_bytes = 0; @@ -654,6 +653,8 @@ heap_vacuum_rel(Relation rel, const VacuumParams params, startreadtime = pgStatBlockReadTime; startwritetime = pgStatBlockWriteTime; } + instr = InstrQueryAlloc(INSTRUMENT_BUFFERS | INSTRUMENT_WAL); + InstrQueryStart(instr); } /* Used for instrumentation and stats report */ @@ -984,14 +985,14 @@ heap_vacuum_rel(Relation rel, const VacuumParams params, { TimestampTz endtime = GetCurrentTimestamp(); + InstrQueryStopFinalize(instr); + if (verbose || params.log_vacuum_min_duration == 0 || TimestampDifferenceExceeds(starttime, endtime, params.log_vacuum_min_duration)) { long secs_dur; int usecs_dur; - WalUsage walusage; - BufferUsage bufferusage; StringInfoData buf; char *msgfmt; int32 diff; @@ -1000,12 +1001,10 @@ heap_vacuum_rel(Relation rel, const VacuumParams params, int64 total_blks_hit; int64 total_blks_read; int64 total_blks_dirtied; + BufferUsage bufferusage = instr->instr.bufusage; + WalUsage walusage = instr->instr.walusage; TimestampDifference(starttime, endtime, &secs_dur, &usecs_dur); - memset(&walusage, 0, sizeof(WalUsage)); - WalUsageAccumDiff(&walusage, &pgWalUsage, &startwalusage); - memset(&bufferusage, 0, sizeof(BufferUsage)); - BufferUsageAccumDiff(&bufferusage, &pgBufferUsage, &startbufferusage); total_blks_hit = bufferusage.shared_blks_hit + bufferusage.local_blks_hit; diff --git a/src/backend/access/nbtree/nbtsort.c b/src/backend/access/nbtree/nbtsort.c index 47a9bda30c9..6a261c8dcbd 100644 --- a/src/backend/access/nbtree/nbtsort.c +++ b/src/backend/access/nbtree/nbtsort.c @@ -1466,8 +1466,8 @@ _bt_begin_parallel(BTBuildState *buildstate, bool isconcurrent, int request) * and PARALLEL_KEY_BUFFER_USAGE. * * If there are no extensions loaded that care, we could skip this. We - * have no way of knowing whether anyone's looking at pgWalUsage or - * pgBufferUsage, so do it unconditionally. + * have no way of knowing whether anyone's looking at instrumentation, so + * do it unconditionally. */ shm_toc_estimate_chunk(&pcxt->estimator, mul_size(sizeof(WalUsage), pcxt->nworkers)); @@ -1753,6 +1753,7 @@ _bt_parallel_build_main(dsm_segment *seg, shm_toc *toc) Relation indexRel; LOCKMODE heapLockmode; LOCKMODE indexLockmode; + QueryInstrumentation *instr; WalUsage *walusage; BufferUsage *bufferusage; int sortmem; @@ -1828,7 +1829,7 @@ _bt_parallel_build_main(dsm_segment *seg, shm_toc *toc) } /* Prepare to track buffer usage during parallel execution */ - InstrStartParallelQuery(); + instr = InstrStartParallelQuery(); /* Perform sorting of spool, and possibly a spool2 */ sortmem = maintenance_work_mem / btshared->scantuplesortstates; @@ -1838,7 +1839,8 @@ _bt_parallel_build_main(dsm_segment *seg, shm_toc *toc) /* Report WAL/buffer usage during parallel execution */ bufferusage = shm_toc_lookup(toc, PARALLEL_KEY_BUFFER_USAGE, false); walusage = shm_toc_lookup(toc, PARALLEL_KEY_WAL_USAGE, false); - InstrEndParallelQuery(&bufferusage[ParallelWorkerNumber], + InstrEndParallelQuery(instr, + &bufferusage[ParallelWorkerNumber], &walusage[ParallelWorkerNumber]); #ifdef BTREE_BUILD_STATS diff --git a/src/backend/commands/analyze.c b/src/backend/commands/analyze.c index eeed91be266..c21b2019eab 100644 --- a/src/backend/commands/analyze.c +++ b/src/backend/commands/analyze.c @@ -309,9 +309,7 @@ do_analyze_rel(Relation onerel, const VacuumParams params, Oid save_userid; int save_sec_context; int save_nestlevel; - WalUsage startwalusage = pgWalUsage; - BufferUsage startbufferusage = pgBufferUsage; - BufferUsage bufferusage; + QueryInstrumentation *instr = NULL; PgStat_Counter startreadtime = 0; PgStat_Counter startwritetime = 0; @@ -362,6 +360,9 @@ do_analyze_rel(Relation onerel, const VacuumParams params, } pg_rusage_init(&ru0); + + instr = InstrQueryAlloc(INSTRUMENT_BUFFERS | INSTRUMENT_WAL); + InstrQueryStart(instr); } /* Used for instrumentation and stats report */ @@ -742,12 +743,13 @@ do_analyze_rel(Relation onerel, const VacuumParams params, { TimestampTz endtime = GetCurrentTimestamp(); + InstrQueryStopFinalize(instr); + if (verbose || params.log_analyze_min_duration == 0 || TimestampDifferenceExceeds(starttime, endtime, params.log_analyze_min_duration)) { long delay_in_ms; - WalUsage walusage; double read_rate = 0; double write_rate = 0; char *msgfmt; @@ -755,18 +757,15 @@ do_analyze_rel(Relation onerel, const VacuumParams params, int64 total_blks_hit; int64 total_blks_read; int64 total_blks_dirtied; - - memset(&bufferusage, 0, sizeof(BufferUsage)); - BufferUsageAccumDiff(&bufferusage, &pgBufferUsage, &startbufferusage); - memset(&walusage, 0, sizeof(WalUsage)); - WalUsageAccumDiff(&walusage, &pgWalUsage, &startwalusage); - - total_blks_hit = bufferusage.shared_blks_hit + - bufferusage.local_blks_hit; - total_blks_read = bufferusage.shared_blks_read + - bufferusage.local_blks_read; - total_blks_dirtied = bufferusage.shared_blks_dirtied + - bufferusage.local_blks_dirtied; + BufferUsage bufusage = instr->instr.bufusage; + WalUsage walusage = instr->instr.walusage; + + total_blks_hit = bufusage.shared_blks_hit + + bufusage.local_blks_hit; + total_blks_read = bufusage.shared_blks_read + + bufusage.local_blks_read; + total_blks_dirtied = bufusage.shared_blks_dirtied + + bufusage.local_blks_dirtied; /* * We do not expect an analyze to take > 25 days and it simplifies diff --git a/src/backend/commands/explain.c b/src/backend/commands/explain.c index e73dc129132..dc5e63955bc 100644 --- a/src/backend/commands/explain.c +++ b/src/backend/commands/explain.c @@ -324,14 +324,16 @@ standard_ExplainOneQuery(Query *query, int cursorOptions, QueryEnvironment *queryEnv) { PlannedStmt *plan; - instr_time planstart, - planduration; - BufferUsage bufusage_start, - bufusage; + QueryInstrumentation *instr = NULL; MemoryContextCounters mem_counters; MemoryContext planner_ctx = NULL; MemoryContext saved_ctx = NULL; + if (es->buffers) + instr = InstrQueryAlloc(INSTRUMENT_TIMER | INSTRUMENT_BUFFERS); + else + instr = InstrQueryAlloc(INSTRUMENT_TIMER); + if (es->memory) { /* @@ -348,15 +350,12 @@ standard_ExplainOneQuery(Query *query, int cursorOptions, saved_ctx = MemoryContextSwitchTo(planner_ctx); } - if (es->buffers) - bufusage_start = pgBufferUsage; - INSTR_TIME_SET_CURRENT(planstart); + InstrQueryStart(instr); /* plan the query */ plan = pg_plan_query(query, queryString, cursorOptions, params, es); - INSTR_TIME_SET_CURRENT(planduration); - INSTR_TIME_SUBTRACT(planduration, planstart); + InstrQueryStopFinalize(instr); if (es->memory) { @@ -364,16 +363,9 @@ standard_ExplainOneQuery(Query *query, int cursorOptions, MemoryContextMemConsumed(planner_ctx, &mem_counters); } - /* calc differences of buffer counters. */ - if (es->buffers) - { - memset(&bufusage, 0, sizeof(BufferUsage)); - BufferUsageAccumDiff(&bufusage, &pgBufferUsage, &bufusage_start); - } - /* run it (if needed) and produce output */ ExplainOnePlan(plan, into, es, queryString, params, queryEnv, - &planduration, (es->buffers ? &bufusage : NULL), + &instr->instr.total, (es->buffers ? &instr->instr.bufusage : NULL), es->memory ? &mem_counters : NULL); } @@ -590,7 +582,12 @@ ExplainOnePlan(PlannedStmt *plannedstmt, IntoClause *into, ExplainState *es, /* grab serialization metrics before we destroy the DestReceiver */ if (es->serialize != EXPLAIN_SERIALIZE_NONE) - serializeMetrics = GetSerializationMetrics(dest); + { + SerializeMetrics *metrics = GetSerializationMetrics(dest); + + if (metrics) + memcpy(&serializeMetrics, metrics, sizeof(SerializeMetrics)); + } /* call the DestReceiver's destroy method even during explain */ dest->rDestroy(dest); @@ -1019,7 +1016,7 @@ ExplainPrintSerialize(ExplainState *es, SerializeMetrics *metrics) ExplainIndentText(es); if (es->timing) appendStringInfo(es->str, "Serialization: time=%.3f ms output=" UINT64_FORMAT "kB format=%s\n", - 1000.0 * INSTR_TIME_GET_DOUBLE(metrics->timeSpent), + 1000.0 * INSTR_TIME_GET_DOUBLE(metrics->instr.total), BYTES_TO_KILOBYTES(metrics->bytesSent), format); else @@ -1027,10 +1024,10 @@ ExplainPrintSerialize(ExplainState *es, SerializeMetrics *metrics) BYTES_TO_KILOBYTES(metrics->bytesSent), format); - if (es->buffers && peek_buffer_usage(es, &metrics->bufferUsage)) + if (es->buffers && peek_buffer_usage(es, &metrics->instr.bufusage)) { es->indent++; - show_buffer_usage(es, &metrics->bufferUsage); + show_buffer_usage(es, &metrics->instr.bufusage); es->indent--; } } @@ -1038,13 +1035,13 @@ ExplainPrintSerialize(ExplainState *es, SerializeMetrics *metrics) { if (es->timing) ExplainPropertyFloat("Time", "ms", - 1000.0 * INSTR_TIME_GET_DOUBLE(metrics->timeSpent), + 1000.0 * INSTR_TIME_GET_DOUBLE(metrics->instr.total), 3, es); ExplainPropertyUInteger("Output Volume", "kB", BYTES_TO_KILOBYTES(metrics->bytesSent), es); ExplainPropertyText("Format", format, es); if (es->buffers) - show_buffer_usage(es, &metrics->bufferUsage); + show_buffer_usage(es, &metrics->instr.bufusage); } ExplainCloseGroup("Serialization", "Serialization", true, es); diff --git a/src/backend/commands/explain_dr.c b/src/backend/commands/explain_dr.c index 3c96061cf32..e1fc723c758 100644 --- a/src/backend/commands/explain_dr.c +++ b/src/backend/commands/explain_dr.c @@ -110,15 +110,11 @@ serializeAnalyzeReceive(TupleTableSlot *slot, DestReceiver *self) MemoryContext oldcontext; StringInfo buf = &myState->buf; int natts = typeinfo->natts; - instr_time start, - end; - BufferUsage instr_start; + Instrumentation *instr = &myState->metrics.instr; /* only measure time, buffers if requested */ - if (myState->es->timing) - INSTR_TIME_SET_CURRENT(start); - if (myState->es->buffers) - instr_start = pgBufferUsage; + if (instr->need_timer || instr->need_stack) + InstrStart(instr); /* Set or update my derived attribute info, if needed */ if (myState->attrinfo != typeinfo || myState->nattrs != natts) @@ -186,18 +182,9 @@ serializeAnalyzeReceive(TupleTableSlot *slot, DestReceiver *self) MemoryContextSwitchTo(oldcontext); MemoryContextReset(myState->tmpcontext); - /* Update timing data */ - if (myState->es->timing) - { - INSTR_TIME_SET_CURRENT(end); - INSTR_TIME_ACCUM_DIFF(myState->metrics.timeSpent, end, start); - } - - /* Update buffer metrics */ - if (myState->es->buffers) - BufferUsageAccumDiff(&myState->metrics.bufferUsage, - &pgBufferUsage, - &instr_start); + /* Stop per-tuple measurement */ + if (instr->need_timer || instr->need_stack) + InstrStop(instr); return true; } @@ -233,9 +220,17 @@ serializeAnalyzeStartup(DestReceiver *self, int operation, TupleDesc typeinfo) /* The output buffer is re-used across rows, as in printtup.c */ initStringInfo(&receiver->buf); - /* Initialize results counters */ + /* Initialize metrics and per-tuple instrumentation */ memset(&receiver->metrics, 0, sizeof(SerializeMetrics)); - INSTR_TIME_SET_ZERO(receiver->metrics.timeSpent); + { + int instrument_options = 0; + + if (receiver->es->timing) + instrument_options |= INSTRUMENT_TIMER; + if (receiver->es->buffers) + instrument_options |= INSTRUMENT_BUFFERS; + InstrInitOptions(&receiver->metrics.instr, instrument_options); + } } /* @@ -246,6 +241,8 @@ serializeAnalyzeShutdown(DestReceiver *self) { SerializeDestReceiver *receiver = (SerializeDestReceiver *) self; + InstrFinalizeChild(&receiver->metrics.instr, instr_stack.current); + if (receiver->finfos) pfree(receiver->finfos); receiver->finfos = NULL; @@ -296,16 +293,18 @@ CreateExplainSerializeDestReceiver(ExplainState *es) * receiver if the subject statement is CREATE TABLE AS. In that * case, return all-zeroes stats. */ -SerializeMetrics +/* + * GetSerializationMetrics - get serialization metrics + * + * Returns a pointer to the SerializeMetrics inside the dest receiver, + * or NULL if the receiver is not a SerializeDestReceiver (e.g. an IntoRel + * receiver for CREATE TABLE AS). + */ +SerializeMetrics * GetSerializationMetrics(DestReceiver *dest) { - SerializeMetrics empty; - if (dest->mydest == DestExplainSerialize) - return ((SerializeDestReceiver *) dest)->metrics; - - memset(&empty, 0, sizeof(SerializeMetrics)); - INSTR_TIME_SET_ZERO(empty.timeSpent); + return &((SerializeDestReceiver *) dest)->metrics; - return empty; + return NULL; } diff --git a/src/backend/commands/prepare.c b/src/backend/commands/prepare.c index 876aad2100a..f7e158e4dd9 100644 --- a/src/backend/commands/prepare.c +++ b/src/backend/commands/prepare.c @@ -580,13 +580,16 @@ ExplainExecuteQuery(ExecuteStmt *execstmt, IntoClause *into, ExplainState *es, ListCell *p; ParamListInfo paramLI = NULL; EState *estate = NULL; - instr_time planstart; - instr_time planduration; - BufferUsage bufusage_start, - bufusage; + QueryInstrumentation *instr = NULL; MemoryContextCounters mem_counters; MemoryContext planner_ctx = NULL; MemoryContext saved_ctx = NULL; + int instrument_options = INSTRUMENT_TIMER; + + if (es->buffers) + instrument_options |= INSTRUMENT_BUFFERS; + + instr = InstrQueryAlloc(instrument_options); if (es->memory) { @@ -598,9 +601,7 @@ ExplainExecuteQuery(ExecuteStmt *execstmt, IntoClause *into, ExplainState *es, saved_ctx = MemoryContextSwitchTo(planner_ctx); } - if (es->buffers) - bufusage_start = pgBufferUsage; - INSTR_TIME_SET_CURRENT(planstart); + InstrQueryStart(instr); /* Look it up in the hash table */ entry = FetchPreparedStatement(execstmt->name, true); @@ -635,8 +636,7 @@ ExplainExecuteQuery(ExecuteStmt *execstmt, IntoClause *into, ExplainState *es, cplan = GetCachedPlan(entry->plansource, paramLI, CurrentResourceOwner, pstate->p_queryEnv); - INSTR_TIME_SET_CURRENT(planduration); - INSTR_TIME_SUBTRACT(planduration, planstart); + InstrQueryStopFinalize(instr); if (es->memory) { @@ -644,13 +644,6 @@ ExplainExecuteQuery(ExecuteStmt *execstmt, IntoClause *into, ExplainState *es, MemoryContextMemConsumed(planner_ctx, &mem_counters); } - /* calc differences of buffer counters. */ - if (es->buffers) - { - memset(&bufusage, 0, sizeof(BufferUsage)); - BufferUsageAccumDiff(&bufusage, &pgBufferUsage, &bufusage_start); - } - plan_list = cplan->stmt_list; /* Explain each query */ @@ -660,7 +653,7 @@ ExplainExecuteQuery(ExecuteStmt *execstmt, IntoClause *into, ExplainState *es, if (pstmt->commandType != CMD_UTILITY) ExplainOnePlan(pstmt, into, es, query_string, paramLI, pstate->p_queryEnv, - &planduration, (es->buffers ? &bufusage : NULL), + &instr->instr.total, (es->buffers ? &instr->instr.bufusage : NULL), es->memory ? &mem_counters : NULL); else ExplainOneUtility(pstmt->utilityStmt, into, es, pstate, paramLI); diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c index c69c12dc014..90ac5ccaacd 100644 --- a/src/backend/commands/tablecmds.c +++ b/src/backend/commands/tablecmds.c @@ -2139,7 +2139,7 @@ ExecuteTruncateGuts(List *explicit_rels, rel, 0, /* dummy rangetable index */ NULL, - 0); + NULL); estate->es_opened_result_relations = lappend(estate->es_opened_result_relations, resultRelInfo); resultRelInfo++; diff --git a/src/backend/commands/trigger.c b/src/backend/commands/trigger.c index 29b80d75143..f2597b917e1 100644 --- a/src/backend/commands/trigger.c +++ b/src/backend/commands/trigger.c @@ -93,6 +93,7 @@ static HeapTuple ExecCallTriggerFunc(TriggerData *trigdata, int tgindx, FmgrInfo *finfo, TriggerInstrumentation *instr, + QueryInstrumentation *qinstr, MemoryContext per_tuple_context); static void AfterTriggerSaveEvent(EState *estate, ResultRelInfo *relinfo, ResultRelInfo *src_partinfo, @@ -2312,6 +2313,7 @@ ExecCallTriggerFunc(TriggerData *trigdata, int tgindx, FmgrInfo *finfo, TriggerInstrumentation *instr, + QueryInstrumentation *qinstr, MemoryContext per_tuple_context) { LOCAL_FCINFO(fcinfo, 0); @@ -2346,7 +2348,7 @@ ExecCallTriggerFunc(TriggerData *trigdata, * If doing EXPLAIN ANALYZE, start charging time to this trigger. */ if (instr) - InstrStartTrigger(instr + tgindx); + InstrStartTrigger(qinstr, instr + tgindx); /* * Do the function evaluation in the per-tuple memory context, so that @@ -2441,6 +2443,7 @@ ExecBSInsertTriggers(EState *estate, ResultRelInfo *relinfo) i, relinfo->ri_TrigFunctions, relinfo->ri_TrigInstrument, + estate->es_instrument, GetPerTupleMemoryContext(estate)); if (newtuple) @@ -2502,6 +2505,7 @@ ExecBRInsertTriggers(EState *estate, ResultRelInfo *relinfo, i, relinfo->ri_TrigFunctions, relinfo->ri_TrigInstrument, + estate->es_instrument, GetPerTupleMemoryContext(estate)); if (newtuple == NULL) { @@ -2606,6 +2610,7 @@ ExecIRInsertTriggers(EState *estate, ResultRelInfo *relinfo, i, relinfo->ri_TrigFunctions, relinfo->ri_TrigInstrument, + estate->es_instrument, GetPerTupleMemoryContext(estate)); if (newtuple == NULL) { @@ -2670,6 +2675,7 @@ ExecBSDeleteTriggers(EState *estate, ResultRelInfo *relinfo) i, relinfo->ri_TrigFunctions, relinfo->ri_TrigInstrument, + estate->es_instrument, GetPerTupleMemoryContext(estate)); if (newtuple) @@ -2780,6 +2786,7 @@ ExecBRDeleteTriggers(EState *estate, EPQState *epqstate, i, relinfo->ri_TrigFunctions, relinfo->ri_TrigInstrument, + estate->es_instrument, GetPerTupleMemoryContext(estate)); if (newtuple == NULL) { @@ -2884,6 +2891,7 @@ ExecIRDeleteTriggers(EState *estate, ResultRelInfo *relinfo, i, relinfo->ri_TrigFunctions, relinfo->ri_TrigInstrument, + estate->es_instrument, GetPerTupleMemoryContext(estate)); if (rettuple == NULL) return false; /* Delete was suppressed */ @@ -2942,6 +2950,7 @@ ExecBSUpdateTriggers(EState *estate, ResultRelInfo *relinfo) i, relinfo->ri_TrigFunctions, relinfo->ri_TrigInstrument, + estate->es_instrument, GetPerTupleMemoryContext(estate)); if (newtuple) @@ -3094,6 +3103,7 @@ ExecBRUpdateTriggers(EState *estate, EPQState *epqstate, i, relinfo->ri_TrigFunctions, relinfo->ri_TrigInstrument, + estate->es_instrument, GetPerTupleMemoryContext(estate)); if (newtuple == NULL) @@ -3258,6 +3268,7 @@ ExecIRUpdateTriggers(EState *estate, ResultRelInfo *relinfo, i, relinfo->ri_TrigFunctions, relinfo->ri_TrigInstrument, + estate->es_instrument, GetPerTupleMemoryContext(estate)); if (newtuple == NULL) { @@ -3316,6 +3327,7 @@ ExecBSTruncateTriggers(EState *estate, ResultRelInfo *relinfo) i, relinfo->ri_TrigFunctions, relinfo->ri_TrigInstrument, + estate->es_instrument, GetPerTupleMemoryContext(estate)); if (newtuple) @@ -4373,7 +4385,7 @@ AfterTriggerExecute(EState *estate, * to include time spent re-fetching tuples in the trigger cost. */ if (instr) - InstrStartTrigger(instr + tgindx); + InstrStartTrigger(estate->es_instrument, instr + tgindx); /* * Fetch the required tuple(s). @@ -4561,6 +4573,7 @@ AfterTriggerExecute(EState *estate, tgindx, finfo, NULL, + NULL, per_tuple_context); if (rettuple != NULL && rettuple != LocTriggerData.tg_trigtuple && diff --git a/src/backend/commands/vacuumparallel.c b/src/backend/commands/vacuumparallel.c index 77834b96a21..c330c891c03 100644 --- a/src/backend/commands/vacuumparallel.c +++ b/src/backend/commands/vacuumparallel.c @@ -308,8 +308,8 @@ parallel_vacuum_init(Relation rel, Relation *indrels, int nindexes, * PARALLEL_VACUUM_KEY_BUFFER_USAGE and PARALLEL_VACUUM_KEY_WAL_USAGE. * * If there are no extensions loaded that care, we could skip this. We - * have no way of knowing whether anyone's looking at pgBufferUsage or - * pgWalUsage, so do it unconditionally. + * have no way of knowing whether anyone's looking at instrumentation, so + * do it unconditionally. */ shm_toc_estimate_chunk(&pcxt->estimator, mul_size(sizeof(BufferUsage), pcxt->nworkers)); @@ -1006,6 +1006,7 @@ parallel_vacuum_main(dsm_segment *seg, shm_toc *toc) PVIndStats *indstats; PVShared *shared; TidStore *dead_items; + QueryInstrumentation *instr; BufferUsage *buffer_usage; WalUsage *wal_usage; int nindexes; @@ -1095,7 +1096,7 @@ parallel_vacuum_main(dsm_segment *seg, shm_toc *toc) error_context_stack = &errcallback; /* Prepare to track buffer usage during parallel execution */ - InstrStartParallelQuery(); + instr = InstrStartParallelQuery(); /* Process indexes to perform vacuum/cleanup */ parallel_vacuum_process_safe_indexes(&pvs); @@ -1103,7 +1104,8 @@ parallel_vacuum_main(dsm_segment *seg, shm_toc *toc) /* Report buffer/WAL usage during parallel execution */ buffer_usage = shm_toc_lookup(toc, PARALLEL_VACUUM_KEY_BUFFER_USAGE, false); wal_usage = shm_toc_lookup(toc, PARALLEL_VACUUM_KEY_WAL_USAGE, false); - InstrEndParallelQuery(&buffer_usage[ParallelWorkerNumber], + InstrEndParallelQuery(instr, + &buffer_usage[ParallelWorkerNumber], &wal_usage[ParallelWorkerNumber]); /* Report any remaining cost-based vacuum delay time */ diff --git a/src/backend/executor/README.instrument b/src/backend/executor/README.instrument new file mode 100644 index 00000000000..580fd5d85e0 --- /dev/null +++ b/src/backend/executor/README.instrument @@ -0,0 +1,227 @@ +src/backend/executor/README.instrument + +Instrumentation +=============== + +The instrumentation subsystem measures time, buffer usage and WAL activity +during query execution and other similar activities. It is used by +EXPLAIN ANALYZE, pg_stat_statements, and other consumers that need +activity and/or timing metrics over a section of code. + +The design has two central goals: + +* Make it cheap to measure activity in a section of code, even when + that section is called many times and the aggregate is what is used + (as is the case with per-node instrumentation in the executor) + +* Ensure nested instrumentation accurately measures activity/timing, + and counter updates from activity get written to the currently + active instrumentation and accumulated upward to parent nodes when + finalized, considering aborts due to errors. + +The key data structures are defined in src/include/executor/instrument.h +and the implementation lives in src/backend/executor/instrument.c. + + +Instrumentation Options +----------------------- + +Callers specify what to measure with a bitmask of InstrumentOption flags: + + INSTRUMENT_ROWS -- row counts only (used with NodeInstrumentation) + INSTRUMENT_TIMER -- wall-clock timing and row counts + INSTRUMENT_BUFFERS -- buffer hit/read/dirtied/written counts and I/O time + INSTRUMENT_WAL -- WAL records, FPI, bytes + +INSTRUMENT_BUFFERS and INSTRUMENT_WAL utilize the instrumentation stack +(described below) for efficient handling of counter values. + + +Struct Hierarchy +---------------- + +There are four instrumentation structs, each specialized for a different +scope: + +Instrumentation Base struct. Holds timing and buffer/WAL counters. + +QueryInstrumentation Extends Instrumentation for query-level tracking. When + stack-based tracking is enabled, it owns a dedicated + MemoryContext and uses the ResourceOwner mechanism for + abort cleanup. + +NodeInstrumentation Extends Instrumentation for per-plan-node statistics + (startup time, tuple counts, loop counts, etc). + +TriggerInstrumentation Extends Instrumentation with a firing count. + + +Stack-based instrumentation +=========================== + +For tracking WAL or buffer usage counters, the specialized stack-based +instrumentation is used. + +At all times, there is a stack that tracks which Instrumentation is currently +active. The stack is represented by instr_stack, a per-backend global +that holds a dynamic array of Instrumentation pointers. The field +instr_stack.current always points to the current stack entry that should +be updated when activity occurs. When the stack array is empty, the +current stack points to instr_top. + +For example, if a backend has two portals open, the overall nesting of +Instrumentation and their respective InstrStart/InstrStop calls creates a +tree-like structure like this: + + Session (instr_top) + | + +-- Query A (QueryInstrumentation) + | | + | +-- NestLoop (NodeInstrumentation) + | | + | +-- Seq Scan A (NodeInstrumentation) + | +-- Seq Scan B (NodeInstrumentation) + | + +-- Query B (QueryInstrumentation) + | + +-- Seq Scan C (NodeInstrumentation) + +While executing Seq Scan B, the stack looks like: + + instr_top (implicit bottom, not in the entries array) + 0: Query A + 1: NestLoop + 2: Seq Scan B <-- instr_stack.current + +When no query is running, the stack is empty (stack_size == 0) and +instr_stack.current points to instr_top. + +Any buffer or WAL counter update (via the INSTR_BUFUSAGE_* and +INSTR_WALUSAGE_* macros in the buffer manager, WAL insertion code, etc.) +writes directly into instr_stack.current. Each instrumentation node starts +zeroed, so the values it accumulates while on top of the stack represent +exactly the activity that occurred during that time. + +Every Instrumentation node has a target, or parent, it will be accumulated +into, which is typically the Instrumentation that was the current stack +entry when it was created. + +For example, when Seq Scan A gets finalized in regular execution via ExecutorFinish, +its instrumentation data gets added to the immediate parent in +the execution tree, the NestLoop, which will then get added to Query A's +QueryInstrumentation, which then accumulates to the parent. + +While we can typically think of this as a tree, the NodeInstrumentation +underneath a particular QueryInstrumentation could behave differently -- +for example, it could propagate directly to the QueryInstrumentation, in +order to not show cumulative numbers in EXPLAIN ANALYZE. + +Note these relationships are partially implicit, especially when it comes +to NodeInstrumentation. Each QueryInstrumentation maintains a list of its +unfinalized child nodes. The parent of a QueryInstrumentation itself is +determined by the stack (see below): when a query is finalized or cleaned +up on abort, its counters are accumulated to whatever entry is then current +on the stack, which is typically instr_top. + + +Finalization and Abort Safety +============================= + +Finalization is the process of rolling up a node's buffer/WAL counters to +its parent. In normal execution, nodes are pushed onto the stack when they +start and popped when they stop; at finalization time their accumulated +counters are added to the parent. + +Due to the use of longjmp for error handling, functions can exit abruptly +without executing their normal cleanup code. On abort, two things need +to happen: + +1. Reset the stack to the appropriate level. This ensures that we don't + later try to update counters on a freed stack entry. We also need to + ensure that the stack entry that was current before a particular + Instrumentation started, is current again after it stops. + +2. Finalize all affected Instrumentation nodes, rolling up their counters + to the highest surviving Instrumentation, so that data is not lost. + +For example, if Seq Scan B aborts while the stack is: + + instr_top (implicit bottom) + 0: Query A + 1: NestLoop + 2: Seq Scan B + +The abort handler for Query A accumulates all unfinalized children (Seq +Scan A, Seq Scan B, NestLoop) directly into Query A's counters, then +unwinds the stack and accumulates Query A's counters to instr_top. + +Note that on abort the children do not accumulate through each other (Seq +Scan B -> NestLoop -> Query A); they all accumulate directly to their +parent QueryInstrumentation. This means the order in which children are +released does not matter -- important because ResourceOwner cleanup does +not guarantee a particular release order. The per-node breakdown is lost, +but the query-level total is what survives the abort. + +If multiple QueryInstrumentations are active on the stack (e.g. nested +portals), each one's abort handler uses InstrStopFinalize to unwind to +whichever entry is higher up, so they compose correctly regardless of +release order. + +There are two mechanisms for achieving abort safety: + +Resource Owner (QueryInstrumentation) +------------------------------------- + +QueryInstrumentation registers with the current ResourceOwner at start. +On transaction abort, the resource owner system calls the release callback, +which walks unfinalized child entries, accumulates their data, unwinds the +stack, and destroys the dedicated memory context (freeing the +QueryInstrumentation and all child allocations as a unit). + +This is the recommended approach when the instrumented code already has an +appropriate resource owner (e.g. it runs inside a portal). The query +executor uses this path. + +PG_FINALLY (base Instrumentation) +---------------------------------- + +When no suitable resource owner exists, or when the caller wants to inspect +the instrumentation data even after an error, the base Instrumentation can +be used with a PG_TRY/PG_FINALLY block that calls InstrStopFinalize(). + +Both mechanisms add overhead, so neither is suitable for high-frequency +instrumentation like per-node measurements in the executor. Instead, +plan node and trigger children rely on their parent QueryInstrumentation +for abort safety: they are allocated in the parent's memory context and +registered in its unfinalized-entries list, so the parent's abort handler +recovers their data automatically. In normal execution, children are +finalized explicitly by the caller. + +Parallel Query +-------------- + +Parallel workers get their own QueryInstrumentation so they can measure +buffer and WAL activity independently, then copy the totals into shared +memory at shutdown. The leader accumulates these into its own stack. + +When per-node instrumentation is active, parallel workers skip per-node +finalization at shutdown to avoid double-counting; the per-node data is +aggregated separately through InstrAggNode(). + + +Memory Handling +=============== + +Instrumentation objects that use the stack must survive until finalization +runs, including the abort case. To ensure this, QueryInstrumentation +creates a dedicated "Instrumentation" MemoryContext (instr_cxt) as a child +of TopMemoryContext. All child instrumentation (nodes, triggers) should be +allocated in this context. + +On successful completion, instr_cxt is reparented to CurrentMemoryContext +so its lifetime is tied to the caller's context. On abort, the +ResourceOwner cleanup frees it after accumulating the instrumentation data +to the current stack entry after resetting the stack. + +When the stack is not needed (timer/rows only), Instrumentation allocations +happen in CurrentMemoryContext instead of TopMemoryContext. diff --git a/src/backend/executor/execMain.c b/src/backend/executor/execMain.c index 1b950040597..5366c1e801c 100644 --- a/src/backend/executor/execMain.c +++ b/src/backend/executor/execMain.c @@ -78,6 +78,7 @@ ExecutorCheckPerms_hook_type ExecutorCheckPerms_hook = NULL; /* decls for local routines only used within this module */ static void InitPlan(QueryDesc *queryDesc, int eflags); static void CheckValidRowMarkRel(Relation rel, RowMarkType markType); +static void ExecFinalizeTriggerInstrumentation(EState *estate); static void ExecPostprocessPlan(EState *estate); static void ExecEndPlan(PlanState *planstate, EState *estate); static void ExecutePlan(QueryDesc *queryDesc, @@ -247,9 +248,19 @@ standard_ExecutorStart(QueryDesc *queryDesc, int eflags) estate->es_snapshot = RegisterSnapshot(queryDesc->snapshot); estate->es_crosscheck_snapshot = RegisterSnapshot(queryDesc->crosscheck_snapshot); estate->es_top_eflags = eflags; - estate->es_instrument = queryDesc->instrument_options; estate->es_jit_flags = queryDesc->plannedstmt->jitFlags; + /* + * Set up query-level instrumentation if needed. We do this before + * InitPlan so that node and trigger instrumentation can be allocated + * within the query's dedicated instrumentation memory context. + */ + if (!queryDesc->totaltime && queryDesc->instrument_options) + { + queryDesc->totaltime = InstrQueryAlloc(queryDesc->instrument_options); + estate->es_instrument = queryDesc->totaltime; + } + /* * Set up an AFTER-trigger statement context, unless told not to, or * unless it's EXPLAIN-only mode (when ExecutorFinish won't be called). @@ -331,9 +342,21 @@ standard_ExecutorRun(QueryDesc *queryDesc, */ oldcontext = MemoryContextSwitchTo(estate->es_query_cxt); - /* Allow instrumentation of Executor overall runtime */ + /* Start up instrumentation for this execution run */ if (queryDesc->totaltime) - InstrStart(queryDesc->totaltime); + { + InstrQueryStart(queryDesc->totaltime); + + /* + * Remember all node entries for abort recovery. We do this once here + * after InstrQueryStart has pushed the parent stack entry. + */ + if (estate->es_instrument && + estate->es_instrument->instr.need_stack && + !queryDesc->already_executed) + ExecRememberNodeInstrumentation(queryDesc->planstate, + queryDesc->totaltime); + } /* * extract information from the query descriptor and the query feature. @@ -385,7 +408,7 @@ standard_ExecutorRun(QueryDesc *queryDesc, dest->rShutdown(dest); if (queryDesc->totaltime) - InstrStop(queryDesc->totaltime); + InstrQueryStop(queryDesc->totaltime); MemoryContextSwitchTo(oldcontext); } @@ -435,7 +458,7 @@ standard_ExecutorFinish(QueryDesc *queryDesc) /* Allow instrumentation of Executor overall runtime */ if (queryDesc->totaltime) - InstrStart(queryDesc->totaltime); + InstrQueryStart(queryDesc->totaltime); /* Run ModifyTable nodes to completion */ ExecPostprocessPlan(estate); @@ -444,8 +467,26 @@ standard_ExecutorFinish(QueryDesc *queryDesc) if (!(estate->es_top_eflags & EXEC_FLAG_SKIP_TRIGGERS)) AfterTriggerEndQuery(estate); + /* + * Accumulate per-node and trigger statistics to their respective parent + * instrumentation stacks. + * + * We skip this in parallel workers because their per-node stats are + * reported individually via ExecParallelReportInstrumentation, and the + * leader's own ExecFinalizeNodeInstrumentation handles propagation. If + * we accumulated here, the leader would double-count: worker parent nodes + * would already include their children's stats, and then the leader's + * accumulation would add the children again. + */ + if (queryDesc->totaltime && estate->es_instrument && !IsParallelWorker()) + { + ExecFinalizeNodeInstrumentation(queryDesc->planstate); + + ExecFinalizeTriggerInstrumentation(estate); + } + if (queryDesc->totaltime) - InstrStop(queryDesc->totaltime); + InstrQueryStopFinalize(queryDesc->totaltime); MemoryContextSwitchTo(oldcontext); @@ -1263,7 +1304,7 @@ InitResultRelInfo(ResultRelInfo *resultRelInfo, Relation resultRelationDesc, Index resultRelationIndex, ResultRelInfo *partition_root_rri, - int instrument_options) + QueryInstrumentation *qinstr) { MemSet(resultRelInfo, 0, sizeof(ResultRelInfo)); resultRelInfo->type = T_ResultRelInfo; @@ -1284,8 +1325,8 @@ InitResultRelInfo(ResultRelInfo *resultRelInfo, palloc0_array(FmgrInfo, n); resultRelInfo->ri_TrigWhenExprs = (ExprState **) palloc0_array(ExprState *, n); - if (instrument_options) - resultRelInfo->ri_TrigInstrument = InstrAllocTrigger(n, instrument_options); + if (qinstr) + resultRelInfo->ri_TrigInstrument = InstrAllocTrigger(qinstr, n); } else { @@ -1499,6 +1540,30 @@ ExecGetAncestorResultRels(EState *estate, ResultRelInfo *resultRelInfo) return resultRelInfo->ri_ancestorResultRels; } +static void +ExecFinalizeTriggerInstrumentation(EState *estate) +{ + List *rels = NIL; + + rels = list_concat(rels, estate->es_tuple_routing_result_relations); + rels = list_concat(rels, estate->es_opened_result_relations); + rels = list_concat(rels, estate->es_trig_target_relations); + + foreach_node(ResultRelInfo, rInfo, rels) + { + TriggerInstrumentation *ti = rInfo->ri_TrigInstrument; + + if (ti == NULL || rInfo->ri_TrigDesc == NULL) + continue; + + for (int nt = 0; nt < rInfo->ri_TrigDesc->numtriggers; nt++) + { + if (ti[nt].instr.need_stack) + InstrAccumStack(&estate->es_instrument->instr, &ti[nt].instr); + } + } +} + /* ---------------------------------------------------------------- * ExecPostprocessPlan * diff --git a/src/backend/executor/execParallel.c b/src/backend/executor/execParallel.c index c153d5c1c3b..0b18a05c434 100644 --- a/src/backend/executor/execParallel.c +++ b/src/backend/executor/execParallel.c @@ -694,7 +694,7 @@ ExecInitParallelPlan(PlanState *planstate, EState *estate, * * If EXPLAIN is not in use and there are no extensions loaded that care, * we could skip this. But we have no way of knowing whether anyone's - * looking at pgBufferUsage, so do it unconditionally. + * looking at instrumentation, so do it unconditionally. */ shm_toc_estimate_chunk(&pcxt->estimator, mul_size(sizeof(BufferUsage), pcxt->nworkers)); @@ -819,13 +819,13 @@ ExecInitParallelPlan(PlanState *planstate, EState *estate, int i; instrumentation = shm_toc_allocate(pcxt->toc, instrumentation_len); - instrumentation->instrument_options = estate->es_instrument; + instrumentation->instrument_options = estate->es_instrument->instrument_options; instrumentation->instrument_offset = instrument_offset; instrumentation->num_workers = nworkers; instrumentation->num_plan_nodes = e.nnodes; instrument = GetInstrumentationArray(instrumentation); for (i = 0; i < nworkers * e.nnodes; ++i) - InstrInitNode(&instrument[i], estate->es_instrument); + InstrInitNode(&instrument[i], estate->es_instrument->instrument_options); shm_toc_insert(pcxt->toc, PARALLEL_KEY_INSTRUMENTATION, instrumentation); pei->instrumentation = instrumentation; @@ -1075,14 +1075,28 @@ ExecParallelRetrieveInstrumentation(PlanState *planstate, instrument = GetInstrumentationArray(instrumentation); instrument += i * instrumentation->num_workers; for (n = 0; n < instrumentation->num_workers; ++n) + { InstrAggNode(planstate->instrument, &instrument[n]); + /* + * Also add worker WAL usage to the global pgWalUsage counter. + * + * When per-node instrumentation is active, parallel workers skip + * ExecFinalizeNodeInstrumentation (to avoid double-counting in + * EXPLAIN), so per-node WAL activity is not rolled up into the + * query-level stats that InstrAccumParallelQuery receives. Without + * this, pgWalUsage would under-report WAL generated by parallel + * workers when instrumentation is active. + */ + WalUsageAdd(&pgWalUsage, &instrument[n].instr.walusage); + } + /* * Also store the per-worker detail. * - * Worker instrumentation should be allocated in the same context as the - * regular instrumentation information, which is the per-query context. - * Switch into per-query memory context. + * Ensure worker instrumentation is allocated in the per-query context. We + * don't need to place this in the instrumentation context since no more + * stack-based instrumentation work is being done. */ oldcontext = MemoryContextSwitchTo(planstate->state->es_query_cxt); ibytes = mul_size(instrumentation->num_workers, sizeof(NodeInstrumentation)); @@ -1456,6 +1470,7 @@ void ParallelQueryMain(dsm_segment *seg, shm_toc *toc) { FixedParallelExecutorState *fpes; + QueryInstrumentation *instr; BufferUsage *buffer_usage; WalUsage *wal_usage; DestReceiver *receiver; @@ -1516,7 +1531,7 @@ ParallelQueryMain(dsm_segment *seg, shm_toc *toc) * leader, which also doesn't count buffer accesses and WAL activity that * occur during executor startup. */ - InstrStartParallelQuery(); + instr = InstrStartParallelQuery(); /* * Run the plan. If we specified a tuple bound, be careful not to demand @@ -1532,7 +1547,8 @@ ParallelQueryMain(dsm_segment *seg, shm_toc *toc) /* Report buffer/WAL usage during parallel execution. */ buffer_usage = shm_toc_lookup(toc, PARALLEL_KEY_BUFFER_USAGE, false); wal_usage = shm_toc_lookup(toc, PARALLEL_KEY_WAL_USAGE, false); - InstrEndParallelQuery(&buffer_usage[ParallelWorkerNumber], + InstrEndParallelQuery(instr, + &buffer_usage[ParallelWorkerNumber], &wal_usage[ParallelWorkerNumber]); /* Report instrumentation data if any instrumentation options are set. */ diff --git a/src/backend/executor/execPartition.c b/src/backend/executor/execPartition.c index d96d4f9947b..6f2909a1bc3 100644 --- a/src/backend/executor/execPartition.c +++ b/src/backend/executor/execPartition.c @@ -1381,7 +1381,7 @@ ExecInitPartitionDispatchInfo(EState *estate, { ResultRelInfo *rri = makeNode(ResultRelInfo); - InitResultRelInfo(rri, rel, 0, rootResultRelInfo, 0); + InitResultRelInfo(rri, rel, 0, rootResultRelInfo, NULL); proute->nonleaf_partitions[dispatchidx] = rri; } else diff --git a/src/backend/executor/execProcnode.c b/src/backend/executor/execProcnode.c index 132fe37ef60..21ad1b04a57 100644 --- a/src/backend/executor/execProcnode.c +++ b/src/backend/executor/execProcnode.c @@ -123,6 +123,8 @@ static TupleTableSlot *ExecProcNodeFirst(PlanState *node); static TupleTableSlot *ExecProcNodeInstr(PlanState *node); static bool ExecShutdownNode_walker(PlanState *node, void *context); +static bool ExecRememberNodeInstrumentation_walker(PlanState *node, void *context); +static bool ExecFinalizeNodeInstrumentation_walker(PlanState *node, void *context); /* ------------------------------------------------------------------------ @@ -788,10 +790,10 @@ ExecShutdownNode_walker(PlanState *node, void *context) * at least once already. We don't expect much CPU consumption during * node shutdown, but in the case of Gather or Gather Merge, we may shut * down workers at this stage. If so, their buffer usage will get - * propagated into pgBufferUsage at this point, and we want to make sure - * that it gets associated with the Gather node. We skip this if the node - * has never been executed, so as to avoid incorrectly making it appear - * that it has. + * propagated into the current instrumentation stack entry at this point, + * and we want to make sure that it gets associated with the Gather node. + * We skip this if the node has never been executed, so as to avoid + * incorrectly making it appear that it has. */ if (node->instrument && node->instrument->running) InstrStartNode(node->instrument); @@ -829,6 +831,80 @@ ExecShutdownNode_walker(PlanState *node, void *context) return false; } +/* + * ExecRememberNodeInstrumentation + * + * Register all per-node instrumentation entries as unfinalized children of + * the executor's instrumentation. This is needed for abort recovery: if the + * executor aborts, we need to walk each per-node entry to recover buffer/WAL + * data from nodes that never got finalized, that someone might be interested + * in as an aggregate. + */ +void +ExecRememberNodeInstrumentation(PlanState *node, QueryInstrumentation *parent) +{ + (void) ExecRememberNodeInstrumentation_walker(node, parent); +} + +static bool +ExecRememberNodeInstrumentation_walker(PlanState *node, void *context) +{ + QueryInstrumentation *parent = (QueryInstrumentation *) context; + + Assert(parent != NULL); + + if (node == NULL) + return false; + + if (node->instrument) + InstrQueryRememberChild(parent, &node->instrument->instr); + + return planstate_tree_walker(node, ExecRememberNodeInstrumentation_walker, context); +} + +/* + * ExecFinalizeNodeInstrumentation + * + * Accumulate instrumentation stats from all execution nodes to their respective + * parents (or the original parent instrumentation). + * + * This must run after the cleanup done by ExecShutdownNode, and not rely on any + * resources cleaned up by it. We also expect shutdown actions to have occurred, + * e.g. parallel worker instrumentation to have been added to the leader. + */ +void +ExecFinalizeNodeInstrumentation(PlanState *node) +{ + (void) ExecFinalizeNodeInstrumentation_walker(node, instr_stack.current); +} + +static bool +ExecFinalizeNodeInstrumentation_walker(PlanState *node, void *context) +{ + Instrumentation *parent = (Instrumentation *) context; + + Assert(parent != NULL); + + if (node == NULL) + return false; + + /* + * Recurse into children first (bottom-up accumulation), passing our + * instrumentation as the parent context. This ensures children can + * accumulate to us even if they were never executed by the leader (e.g. + * nodes beneath Gather that only workers ran). + */ + planstate_tree_walker(node, ExecFinalizeNodeInstrumentation_walker, + node->instrument ? &node->instrument->instr : parent); + + if (!node->instrument) + return false; + + InstrFinalizeChild(&node->instrument->instr, parent); + + return false; +} + /* * ExecSetTupleBound * diff --git a/src/backend/executor/execUtils.c b/src/backend/executor/execUtils.c index 9886ab06b69..0da0ff6b339 100644 --- a/src/backend/executor/execUtils.c +++ b/src/backend/executor/execUtils.c @@ -150,7 +150,7 @@ CreateExecutorState(void) estate->es_total_processed = 0; estate->es_top_eflags = 0; - estate->es_instrument = 0; + estate->es_instrument = NULL; estate->es_finished = false; estate->es_exprcontexts = NIL; @@ -227,6 +227,15 @@ FreeExecutorState(EState *estate) estate->es_partition_directory = NULL; } + /* + * Make sure the instrumentation context gets freed. This usually gets + * re-parented under the per-query context in InstrQueryStopFinalize, but + * that won't happen during EXPLAIN (BUFFERS) since ExecutorFinish never + * gets called, so we would otherwise leak it in TopMemoryContext. + */ + if (estate->es_instrument && estate->es_instrument->instr.need_stack) + MemoryContextDelete(estate->es_instrument->instr_cxt); + /* * Free the per-query memory context, thereby releasing all working * memory, including the EState node itself. diff --git a/src/backend/executor/instrument.c b/src/backend/executor/instrument.c index bc551f95a08..6892706a83a 100644 --- a/src/backend/executor/instrument.c +++ b/src/backend/executor/instrument.c @@ -16,30 +16,46 @@ #include #include "executor/instrument.h" +#include "utils/memutils.h" +#include "utils/resowner.h" -BufferUsage pgBufferUsage; -static BufferUsage save_pgBufferUsage; WalUsage pgWalUsage; -static WalUsage save_pgWalUsage; +Instrumentation instr_top; +InstrStackState instr_stack = {0, 0, NULL, &instr_top}; -static void BufferUsageAdd(BufferUsage *dst, const BufferUsage *add); -static void WalUsageAdd(WalUsage *dst, WalUsage *add); +void +InstrStackGrow(void) +{ + int space = instr_stack.stack_space; + + if (instr_stack.entries == NULL) + { + space = 10; /* Allocate sufficient initial space for + * typical activity */ + instr_stack.entries = MemoryContextAlloc(TopMemoryContext, + sizeof(Instrumentation *) * space); + } + else + { + space *= 2; + instr_stack.entries = repalloc_array(instr_stack.entries, Instrumentation *, space); + } + /* Update stack space after allocation succeeded to protect against OOMs */ + instr_stack.stack_space = space; +} /* General purpose instrumentation handling */ -Instrumentation * -InstrAlloc(int instrument_options) +static inline bool +InstrNeedStack(int instrument_options) { - Instrumentation *instr = palloc0(sizeof(Instrumentation)); - InstrInitOptions(instr, instrument_options); - return instr; + return (instrument_options & (INSTRUMENT_BUFFERS | INSTRUMENT_WAL)) != 0; } void InstrInitOptions(Instrumentation *instr, int instrument_options) { - instr->need_bufusage = (instrument_options & INSTRUMENT_BUFFERS) != 0; - instr->need_walusage = (instrument_options & INSTRUMENT_WAL) != 0; + instr->need_stack = InstrNeedStack(instrument_options); instr->need_timer = (instrument_options & INSTRUMENT_TIMER) != 0; } @@ -54,50 +70,295 @@ InstrStart(Instrumentation *instr) INSTR_TIME_SET_CURRENT(instr->starttime); } - /* save buffer usage totals at node entry, if needed */ - if (instr->need_bufusage) - instr->bufusage_start = pgBufferUsage; + if (instr->need_stack) + InstrPushStack(instr); +} + +static void +InstrStopTimer(Instrumentation *instr) +{ + instr_time endtime; - if (instr->need_walusage) - instr->walusage_start = pgWalUsage; + /* let's update the time only if the timer was requested */ + if (INSTR_TIME_IS_ZERO(instr->starttime)) + elog(ERROR, "InstrStop called without start"); + + INSTR_TIME_SET_CURRENT(endtime); + INSTR_TIME_ACCUM_DIFF(instr->total, endtime, instr->starttime); + + INSTR_TIME_SET_ZERO(instr->starttime); } void InstrStop(Instrumentation *instr) { - instr_time endtime; + if (instr->need_timer) + InstrStopTimer(instr); + + if (instr->need_stack) + InstrPopStack(instr); +} + +/* + * Stops instrumentation, finalizes the stack entry and accumulates to its parent. + * + * Note that this intentionally allows passing a stack that is not the current + * top, as can happen with PG_FINALLY, or resource owners, which don't have a + * guaranteed cleanup order. + * + * We are careful here to achieve two goals: + * + * 1) Reset the stack to the parent of whichever of the released stack entries + * has the lowest index + * 2) Accumulate all instrumentation to the currently active instrumentation, + * so that callers get a complete picture of activity, even after an abort + */ +void +InstrStopFinalize(Instrumentation *instr) +{ + int idx = -1; + + for (int i = instr_stack.stack_size - 1; i >= 0; i--) + { + if (instr_stack.entries[i] == instr) + { + idx = i; + break; + } + } + + if (idx >= 0) + { + while (instr_stack.stack_size > idx + 1) + instr_stack.stack_size--; + + InstrPopStack(instr); + } - /* let's update the time only if the timer was requested */ if (instr->need_timer) + InstrStopTimer(instr); + + InstrAccumStack(instr_stack.current, instr); +} + +/* + * Finalize child instrumentation by accumulating buffer/WAL usage to the + * provided instrumentation, which may be the current entry, or one the caller + * treats as a parent and will add to the totals later. + * + * Also deletes the unfinalized entry to avoid double counting in an abort + * situation, e.g. during executor finish. + */ +void +InstrFinalizeChild(Instrumentation *instr, Instrumentation *parent) +{ + if (instr->need_stack) { - if (INSTR_TIME_IS_ZERO(instr->starttime)) - elog(ERROR, "InstrStop called without start"); + if (!dlist_node_is_detached(&instr->unfinalized_entry)) + dlist_delete_thoroughly(&instr->unfinalized_entry); - INSTR_TIME_SET_CURRENT(endtime); - INSTR_TIME_ACCUM_DIFF(instr->total, endtime, instr->starttime); + InstrAccumStack(parent, instr); + } +} + + +/* Query instrumentation handling */ + +/* + * Use ResourceOwner mechanism to correctly reset instr_stack on abort. + */ +static void ResOwnerReleaseInstrumentation(Datum res); +static const ResourceOwnerDesc instrumentation_resowner_desc = +{ + .name = "instrumentation", + .release_phase = RESOURCE_RELEASE_AFTER_LOCKS, + .release_priority = RELEASE_PRIO_INSTRUMENTATION, + .ReleaseResource = ResOwnerReleaseInstrumentation, + .DebugPrint = NULL, /* default message is fine */ +}; + +static inline void +ResourceOwnerRememberInstrumentation(ResourceOwner owner, QueryInstrumentation *qinstr) +{ + ResourceOwnerRemember(owner, PointerGetDatum(qinstr), &instrumentation_resowner_desc); +} + +static inline void +ResourceOwnerForgetInstrumentation(ResourceOwner owner, QueryInstrumentation *qinstr) +{ + ResourceOwnerForget(owner, PointerGetDatum(qinstr), &instrumentation_resowner_desc); +} + +static void +ResOwnerReleaseInstrumentation(Datum res) +{ + QueryInstrumentation *qinstr = (QueryInstrumentation *) DatumGetPointer(res); + MemoryContext instr_cxt = qinstr->instr_cxt; + dlist_mutable_iter iter; - INSTR_TIME_SET_ZERO(instr->starttime); + /* Accumulate data from all unfinalized child entries (nodes, triggers) */ + dlist_foreach_modify(iter, &qinstr->unfinalized_entries) + { + Instrumentation *child = dlist_container(Instrumentation, unfinalized_entry, iter.cur); + + InstrAccumStack(&qinstr->instr, child); } - /* Add delta of buffer usage since entry to node's totals */ - if (instr->need_bufusage) - BufferUsageAccumDiff(&instr->bufusage, - &pgBufferUsage, &instr->bufusage_start); + /* Ensure the stack is reset as expected, and we accumulate to the parent */ + InstrStopFinalize(&qinstr->instr); + + /* + * Destroy the dedicated instrumentation context, which frees the + * QueryInstrumentation and all child allocations. + */ + MemoryContextDelete(instr_cxt); +} + +QueryInstrumentation * +InstrQueryAlloc(int instrument_options) +{ + QueryInstrumentation *instr; + MemoryContext instr_cxt; + + /* + * When the instrumentation stack is used, create a dedicated memory + * context for this query's instrumentation allocations. This context is a + * child of TopMemoryContext so it survives transaction abort — + * ResourceOwner release needs to access it. + * + * For simpler cases (timer/rows only), use the current memory context. + * + * All child instrumentation allocations (nodes, triggers, etc) must be + * allocated within this context to ensure correct clean up on abort. + */ + if (InstrNeedStack(instrument_options)) + instr_cxt = AllocSetContextCreate(TopMemoryContext, + "Instrumentation", + ALLOCSET_SMALL_SIZES); + else + instr_cxt = CurrentMemoryContext; - if (instr->need_walusage) - WalUsageAccumDiff(&instr->walusage, - &pgWalUsage, &instr->walusage_start); + instr = MemoryContextAllocZero(instr_cxt, sizeof(QueryInstrumentation)); + instr->instrument_options = instrument_options; + instr->instr_cxt = instr_cxt; + + InstrInitOptions(&instr->instr, instrument_options); + dlist_init(&instr->unfinalized_entries); + + return instr; +} + +void +InstrQueryStart(QueryInstrumentation *qinstr) +{ + InstrStart(&qinstr->instr); + + if (qinstr->instr.need_stack) + { + Assert(CurrentResourceOwner != NULL); + qinstr->owner = CurrentResourceOwner; + + ResourceOwnerEnlarge(qinstr->owner); + ResourceOwnerRememberInstrumentation(qinstr->owner, qinstr); + } +} + +void +InstrQueryStop(QueryInstrumentation *qinstr) +{ + InstrStop(&qinstr->instr); + + if (qinstr->instr.need_stack) + { + Assert(qinstr->owner != NULL); + ResourceOwnerForgetInstrumentation(qinstr->owner, qinstr); + qinstr->owner = NULL; + } +} + +void +InstrQueryStopFinalize(QueryInstrumentation *qinstr) +{ + InstrStopFinalize(&qinstr->instr); + + if (!qinstr->instr.need_stack) + return; + + Assert(qinstr->owner != NULL); + ResourceOwnerForgetInstrumentation(qinstr->owner, qinstr); + qinstr->owner = NULL; + + /* + * Reparent the dedicated instrumentation context under the current memory + * context, so that its lifetime is now tied to the caller's context + * rather than TopMemoryContext. + */ + MemoryContextSetParent(qinstr->instr_cxt, CurrentMemoryContext); +} + +/* + * Register a child Instrumentation entry for abort processing. + * + * On abort, ResOwnerReleaseInstrumentation will walk the parent's list to + * recover buffer/WAL data from entries that were never finalized, in order for + * aggregate totals to be accurate despite the query erroring out. + */ +void +InstrQueryRememberChild(QueryInstrumentation *parent, Instrumentation *child) +{ + if (child->need_stack) + dlist_push_head(&parent->unfinalized_entries, &child->unfinalized_entry); +} + +/* start instrumentation during parallel executor startup */ +QueryInstrumentation * +InstrStartParallelQuery(void) +{ + QueryInstrumentation *qinstr = InstrQueryAlloc(INSTRUMENT_BUFFERS | INSTRUMENT_WAL); + + InstrQueryStart(qinstr); + return qinstr; +} + +/* report usage after parallel executor shutdown */ +void +InstrEndParallelQuery(QueryInstrumentation *qinstr, BufferUsage *bufusage, WalUsage *walusage) +{ + InstrQueryStopFinalize(qinstr); + memcpy(bufusage, &qinstr->instr.bufusage, sizeof(BufferUsage)); + memcpy(walusage, &qinstr->instr.walusage, sizeof(WalUsage)); +} + +/* + * Accumulate work done by parallel workers in the leader's stats. + * + * Note that what gets added here effectively depends on whether per-node + * instrumentation is active. If it's active the parallel worker intentionally + * skips ExecFinalizeNodeInstrumentation on executor shutdown, because it would + * cause double counting. Instead, this only accumulates any extra activity + * outside of nodes. + * + * Otherwise this is responsible for making sure that the complete query + * activity is accumulated. + */ +void +InstrAccumParallelQuery(BufferUsage *bufusage, WalUsage *walusage) +{ + BufferUsageAdd(&instr_stack.current->bufusage, bufusage); + WalUsageAdd(&instr_stack.current->walusage, walusage); + + WalUsageAdd(&pgWalUsage, walusage); } /* Node instrumentation handling */ /* Allocate new node instrumentation structure */ NodeInstrumentation * -InstrAllocNode(int instrument_options, bool async_mode) +InstrAllocNode(QueryInstrumentation *qinstr, bool async_mode) { - NodeInstrumentation *instr = palloc(sizeof(NodeInstrumentation)); + NodeInstrumentation *instr = MemoryContextAlloc(qinstr->instr_cxt, sizeof(NodeInstrumentation)); - InstrInitNode(instr, instrument_options); + InstrInitNode(instr, qinstr->instrument_options); instr->async_mode = async_mode; return instr; @@ -118,6 +379,7 @@ InstrStartNode(NodeInstrumentation *instr) InstrStart(&instr->instr); } + /* Exit from a plan node */ void InstrStopNode(NodeInstrumentation *instr, double nTuples) @@ -147,14 +409,12 @@ InstrStopNode(NodeInstrumentation *instr, double nTuples) INSTR_TIME_SET_ZERO(instr->instr.starttime); } - /* Add delta of buffer usage since entry to node's totals */ - if (instr->instr.need_bufusage) - BufferUsageAccumDiff(&instr->instr.bufusage, - &pgBufferUsage, &instr->instr.bufusage_start); - - if (instr->instr.need_walusage) - WalUsageAccumDiff(&instr->instr.walusage, - &pgWalUsage, &instr->instr.walusage_start); + /* + * Only pop the stack, accumulation runs in + * ExecFinalizeNodeInstrumentation + */ + if (instr->instr.need_stack) + InstrPopStack(&instr->instr); /* Is this the first tuple of this cycle? */ if (!instr->running) @@ -189,8 +449,8 @@ InstrEndLoop(NodeInstrumentation *instr) if (!instr->running) return; - if (!INSTR_TIME_IS_ZERO(instr->instr.starttime)) - elog(ERROR, "InstrEndLoop called on running node"); + /* Ensure InstrNodeStop was called */ + Assert(INSTR_TIME_IS_ZERO(instr->instr.starttime)); /* Accumulate per-cycle statistics into totals */ INSTR_TIME_ADD(instr->startup, instr->firsttuple); @@ -231,67 +491,73 @@ InstrAggNode(NodeInstrumentation *dst, NodeInstrumentation *add) dst->nfiltered2 += add->nfiltered2; /* Add delta of buffer usage since entry to node's totals */ - if (dst->instr.need_bufusage) - BufferUsageAdd(&dst->instr.bufusage, &add->instr.bufusage); - - if (dst->instr.need_walusage) - WalUsageAdd(&dst->instr.walusage, &add->instr.walusage); + if (dst->instr.need_stack) + InstrAccumStack(&dst->instr, &add->instr); } /* Trigger instrumentation handling */ TriggerInstrumentation * -InstrAllocTrigger(int n, int instrument_options) +InstrAllocTrigger(QueryInstrumentation *qinstr, int n) { - TriggerInstrumentation *tginstr = palloc0(n * sizeof(TriggerInstrumentation)); + TriggerInstrumentation *tginstr; int i; + /* + * Allocate in the query's dedicated instrumentation context so all + * instrumentation data is grouped together and cleaned up as a unit. + */ + Assert(qinstr != NULL && qinstr->instr_cxt != NULL); + tginstr = MemoryContextAllocZero(qinstr->instr_cxt, + n * sizeof(TriggerInstrumentation)); + for (i = 0; i < n; i++) - InstrInitOptions(&tginstr[i].instr, instrument_options); + InstrInitOptions(&tginstr[i].instr, qinstr->instrument_options); return tginstr; } void -InstrStartTrigger(TriggerInstrumentation *tginstr) +InstrStartTrigger(QueryInstrumentation *qinstr, TriggerInstrumentation *tginstr) { InstrStart(&tginstr->instr); + + /* + * On first call, register with the parent QueryInstrumentation for abort + * recovery. + */ + if (qinstr && tginstr->instr.need_stack && + dlist_node_is_detached(&tginstr->instr.unfinalized_entry)) + dlist_push_head(&qinstr->unfinalized_entries, + &tginstr->instr.unfinalized_entry); } void InstrStopTrigger(TriggerInstrumentation *tginstr, int firings) { + /* + * This trigger may be called again, so we don't finalize instrumentation + * here. Accumulation to the parent happens at ExecutorFinish through + * ExecFinalizeTriggerInstrumentation. + */ InstrStop(&tginstr->instr); tginstr->firings += firings; } -/* note current values during parallel executor startup */ void -InstrStartParallelQuery(void) +InstrAccumStack(Instrumentation *dst, Instrumentation *add) { - save_pgBufferUsage = pgBufferUsage; - save_pgWalUsage = pgWalUsage; -} + Assert(dst != NULL); + Assert(add != NULL); -/* report usage after parallel executor shutdown */ -void -InstrEndParallelQuery(BufferUsage *bufusage, WalUsage *walusage) -{ - memset(bufusage, 0, sizeof(BufferUsage)); - BufferUsageAccumDiff(bufusage, &pgBufferUsage, &save_pgBufferUsage); - memset(walusage, 0, sizeof(WalUsage)); - WalUsageAccumDiff(walusage, &pgWalUsage, &save_pgWalUsage); -} + if (!add->need_stack) + return; -/* accumulate work done by workers in leader's stats */ -void -InstrAccumParallelQuery(BufferUsage *bufusage, WalUsage *walusage) -{ - BufferUsageAdd(&pgBufferUsage, bufusage); - WalUsageAdd(&pgWalUsage, walusage); + BufferUsageAdd(&dst->bufusage, &add->bufusage); + WalUsageAdd(&dst->walusage, &add->walusage); } /* dst += add */ -static void +void BufferUsageAdd(BufferUsage *dst, const BufferUsage *add) { dst->shared_blks_hit += add->shared_blks_hit; @@ -312,39 +578,9 @@ BufferUsageAdd(BufferUsage *dst, const BufferUsage *add) INSTR_TIME_ADD(dst->temp_blk_write_time, add->temp_blk_write_time); } -/* dst += add - sub */ +/* dst += add */ void -BufferUsageAccumDiff(BufferUsage *dst, - const BufferUsage *add, - const BufferUsage *sub) -{ - dst->shared_blks_hit += add->shared_blks_hit - sub->shared_blks_hit; - dst->shared_blks_read += add->shared_blks_read - sub->shared_blks_read; - dst->shared_blks_dirtied += add->shared_blks_dirtied - sub->shared_blks_dirtied; - dst->shared_blks_written += add->shared_blks_written - sub->shared_blks_written; - dst->local_blks_hit += add->local_blks_hit - sub->local_blks_hit; - dst->local_blks_read += add->local_blks_read - sub->local_blks_read; - dst->local_blks_dirtied += add->local_blks_dirtied - sub->local_blks_dirtied; - dst->local_blks_written += add->local_blks_written - sub->local_blks_written; - dst->temp_blks_read += add->temp_blks_read - sub->temp_blks_read; - dst->temp_blks_written += add->temp_blks_written - sub->temp_blks_written; - INSTR_TIME_ACCUM_DIFF(dst->shared_blk_read_time, - add->shared_blk_read_time, sub->shared_blk_read_time); - INSTR_TIME_ACCUM_DIFF(dst->shared_blk_write_time, - add->shared_blk_write_time, sub->shared_blk_write_time); - INSTR_TIME_ACCUM_DIFF(dst->local_blk_read_time, - add->local_blk_read_time, sub->local_blk_read_time); - INSTR_TIME_ACCUM_DIFF(dst->local_blk_write_time, - add->local_blk_write_time, sub->local_blk_write_time); - INSTR_TIME_ACCUM_DIFF(dst->temp_blk_read_time, - add->temp_blk_read_time, sub->temp_blk_read_time); - INSTR_TIME_ACCUM_DIFF(dst->temp_blk_write_time, - add->temp_blk_write_time, sub->temp_blk_write_time); -} - -/* helper functions for WAL usage accumulation */ -static void -WalUsageAdd(WalUsage *dst, WalUsage *add) +WalUsageAdd(WalUsage *dst, const WalUsage *add) { dst->wal_bytes += add->wal_bytes; dst->wal_records += add->wal_records; diff --git a/src/backend/replication/logical/worker.c b/src/backend/replication/logical/worker.c index 27d398d576d..4f7f097be2f 100644 --- a/src/backend/replication/logical/worker.c +++ b/src/backend/replication/logical/worker.c @@ -903,7 +903,7 @@ create_edata_for_relation(LogicalRepRelMapEntry *rel) * Use Relation opened by logicalrep_rel_open() instead of opening it * again. */ - InitResultRelInfo(resultRelInfo, rel->localrel, 1, NULL, 0); + InitResultRelInfo(resultRelInfo, rel->localrel, 1, NULL, NULL); /* * We put the ResultRelInfo in the es_opened_result_relations list, even diff --git a/src/backend/storage/buffer/bufmgr.c b/src/backend/storage/buffer/bufmgr.c index ce1af4ad563..0a623b68996 100644 --- a/src/backend/storage/buffer/bufmgr.c +++ b/src/backend/storage/buffer/bufmgr.c @@ -1259,9 +1259,9 @@ PinBufferForBlock(Relation rel, if (rel) { /* - * While pgBufferUsage's "read" counter isn't bumped unless we reach - * WaitReadBuffers() (so, not for hits, and not for buffers that are - * zeroed instead), the per-relation stats always count them. + * While the current buffer usage "read" counter isn't bumped unless + * we reach WaitReadBuffers() (so, not for hits, and not for buffers + * that are zeroed instead), the per-relation stats always count them. */ pgstat_count_buffer_read(rel); } diff --git a/src/backend/utils/activity/pgstat_io.c b/src/backend/utils/activity/pgstat_io.c index 9e7a88ec0d0..60400f0c81f 100644 --- a/src/backend/utils/activity/pgstat_io.c +++ b/src/backend/utils/activity/pgstat_io.c @@ -114,9 +114,9 @@ pgstat_prepare_io_time(bool track_io_guc) * pg_stat_database only counts block read and write times, these are done for * IOOP_READ, IOOP_WRITE and IOOP_EXTEND. * - * pgBufferUsage is used for EXPLAIN. pgBufferUsage has write and read stats - * for shared, local and temporary blocks. pg_stat_io does not track the - * activity of temporary blocks, so these are ignored here. + * Executor instrumentation is used for EXPLAIN. Buffer usage tracked there has + * write and read stats for shared, local and temporary blocks. pg_stat_io + * does not track the activity of temporary blocks, so these are ignored here. */ void pgstat_count_io_op_time(IOObject io_object, IOContext io_context, IOOp io_op, diff --git a/src/include/commands/explain_dr.h b/src/include/commands/explain_dr.h index f98eaae1864..fa98d29589f 100644 --- a/src/include/commands/explain_dr.h +++ b/src/include/commands/explain_dr.h @@ -23,11 +23,10 @@ typedef struct ExplainState ExplainState; typedef struct SerializeMetrics { uint64 bytesSent; /* # of bytes serialized */ - instr_time timeSpent; /* time spent serializing */ - BufferUsage bufferUsage; /* buffers accessed during serialization */ + Instrumentation instr; /* per-tuple timing/buffer measurement */ } SerializeMetrics; extern DestReceiver *CreateExplainSerializeDestReceiver(ExplainState *es); -extern SerializeMetrics GetSerializationMetrics(DestReceiver *dest); +extern SerializeMetrics *GetSerializationMetrics(DestReceiver *dest); #endif diff --git a/src/include/executor/execdesc.h b/src/include/executor/execdesc.h index d3a57242844..340029a2034 100644 --- a/src/include/executor/execdesc.h +++ b/src/include/executor/execdesc.h @@ -51,8 +51,8 @@ typedef struct QueryDesc /* This field is set by ExecutePlan */ bool already_executed; /* true if previously executed */ - /* This is always set NULL by the core system, but plugins can change it */ - struct Instrumentation *totaltime; /* total time spent in ExecutorRun */ + /* This field is set by ExecutorRun, or plugins */ + struct QueryInstrumentation *totaltime; /* total time spent in ExecutorRun */ } QueryDesc; /* in pquery.c */ diff --git a/src/include/executor/executor.h b/src/include/executor/executor.h index 07f4b1f7490..f56b13841fb 100644 --- a/src/include/executor/executor.h +++ b/src/include/executor/executor.h @@ -233,6 +233,7 @@ ExecGetJunkAttribute(TupleTableSlot *slot, AttrNumber attno, bool *isNull) /* * prototypes from functions in execMain.c */ +typedef struct QueryInstrumentation QueryInstrumentation; extern void ExecutorStart(QueryDesc *queryDesc, int eflags); extern void standard_ExecutorStart(QueryDesc *queryDesc, int eflags); extern void ExecutorRun(QueryDesc *queryDesc, @@ -254,7 +255,7 @@ extern void InitResultRelInfo(ResultRelInfo *resultRelInfo, Relation resultRelationDesc, Index resultRelationIndex, ResultRelInfo *partition_root_rri, - int instrument_options); + QueryInstrumentation *qinstr); extern ResultRelInfo *ExecGetTriggerResultRel(EState *estate, Oid relid, ResultRelInfo *rootRelInfo); extern List *ExecGetAncestorResultRels(EState *estate, ResultRelInfo *resultRelInfo); @@ -301,6 +302,8 @@ extern void ExecSetExecProcNode(PlanState *node, ExecProcNodeMtd function); extern Node *MultiExecProcNode(PlanState *node); extern void ExecEndNode(PlanState *node); extern void ExecShutdownNode(PlanState *node); +extern void ExecRememberNodeInstrumentation(PlanState *node, QueryInstrumentation *parent); +extern void ExecFinalizeNodeInstrumentation(PlanState *node); extern void ExecSetTupleBound(int64 tuples_needed, PlanState *child_node); diff --git a/src/include/executor/instrument.h b/src/include/executor/instrument.h index d4769f3da7b..f49c3f99cf2 100644 --- a/src/include/executor/instrument.h +++ b/src/include/executor/instrument.h @@ -13,6 +13,7 @@ #ifndef INSTRUMENT_H #define INSTRUMENT_H +#include "lib/ilist.h" #include "portability/instr_time.h" @@ -68,29 +69,91 @@ typedef enum InstrumentOption } InstrumentOption; /* - * General purpose instrumentation that can capture time and WAL/buffer usage + * Instrumentation base class for capturing time and WAL/buffer usage * - * Initialized through InstrAlloc, followed by one or more calls to a pair of - * InstrStart/InstrStop (activity is measured inbetween). + * If used directly: + * - Allocate on the stack and zero initialize the struct + * - Call InstrInitOptions to set instrumentation options + * - Call InstrStart before the activity you want to measure + * - Call InstrStop / InstrStopFinalize after the activity to capture totals + * + * InstrStart/InstrStop may be called multiple times. The last stop call must + * be to InstrStopFinalize to ensure parent stack entries get the accumulated + * totals. If there is risk of transaction aborts you must call + * InstrStopFinalize in a PG_TRY/PG_FINALLY block to avoid corrupting the + * instrumentation stack. + * + * In a query context use QueryInstrumentation instead, which handles aborts + * using the resource owner logic. */ typedef struct Instrumentation { /* Parameters set at creation: */ bool need_timer; /* true if we need timer data */ - bool need_bufusage; /* true if we need buffer usage data */ - bool need_walusage; /* true if we need WAL usage data */ + bool need_stack; /* true if we need WAL/buffer usage data */ /* Internal state keeping: */ instr_time starttime; /* start time of last InstrStart */ - BufferUsage bufusage_start; /* buffer usage at start */ - WalUsage walusage_start; /* WAL usage at start */ /* Accumulated statistics: */ instr_time total; /* total runtime */ BufferUsage bufusage; /* total buffer usage */ WalUsage walusage; /* total WAL usage */ + /* Abort handling: link in parent QueryInstrumentation's unfinalized list */ + dlist_node unfinalized_entry; } Instrumentation; +/* + * Query-related instrumentation tracking. + * + * Usage: + * - Allocate on the heap using InstrQueryAlloc (required for abort handling) + * - Call InstrQueryStart before the activity you want to measure + * - Call InstrQueryStop / InstrQueryStopFinalize afterwards to capture totals + * + * InstrQueryStart/InstrQueryStop may be called multiple times. The last stop + * call must be to InstrQueryStopFinalize to ensure parent stack entries get + * the accumulated totals. + * + * Uses resource owner mechanism for handling aborts, as such, the caller + * *must* not exit out of the top level transaction after having called + * InstrQueryStart, without first calling InstrQueryStop or + * InstrQueryStopFinalize. In the case of a transaction abort, logic equivalent + * to InstrQueryStopFinalize will be called automatically. + */ +struct ResourceOwnerData; +typedef struct QueryInstrumentation +{ + Instrumentation instr; + + /* Original instrument_options flags used to create this instrumentation */ + int instrument_options; + + /* Resource owner used for cleanup for aborts between InstrStart/InstrStop */ + struct ResourceOwnerData *owner; + + /* + * Dedicated memory context for all instrumentation allocations belonging + * to this query (node instrumentation, trigger instrumentation, etc.). + * Initially a child of TopMemoryContext so it survives transaction abort + * for ResourceOwner cleanup, which is then reassigned to the current + * memory context on InstrQueryStopFinalize. + */ + MemoryContext instr_cxt; + + /* + * Child entries that need to be cleaned up on abort, since they are not + * registered as a resource owner themselves. Contains both node and + * trigger instrumentation entries linked via instr.unfinalized_entry. + */ + dlist_head unfinalized_entries; +} QueryInstrumentation; + /* * Specialized instrumentation for per-node execution statistics + * + * Relies on an outer QueryInstrumentation having been set up to handle the + * stack used for WAL/buffer usage statistics, and relies on it for managing + * aborts. Solely intended for the executor and anyone reporting about its + * activities (e.g. EXPLAIN ANALYZE). */ typedef struct NodeInstrumentation { @@ -111,6 +174,10 @@ typedef struct NodeInstrumentation double nfiltered2; /* # of tuples removed by "other" quals */ } NodeInstrumentation; +/* + * Care must be taken with any pointers contained within this struct, as this + * gets copied across processes during parallel query execution. + */ typedef struct WorkerNodeInstrumentation { int num_workers; /* # of structures that follow */ @@ -124,16 +191,102 @@ typedef struct TriggerInstrumentation * was fired */ } TriggerInstrumentation; -extern PGDLLIMPORT BufferUsage pgBufferUsage; +/* + * Dynamic array-based stack for tracking current WAL/buffer usage context. + * + * When the stack is empty, 'current' points to instr_top which accumulates + * session-level totals. + */ +typedef struct InstrStackState +{ + int stack_space; /* allocated capacity of entries array */ + int stack_size; /* current number of entries */ + + Instrumentation **entries; /* dynamic array of pointers */ + Instrumentation *current; /* top of stack, or &instr_top when empty */ +} InstrStackState; + extern PGDLLIMPORT WalUsage pgWalUsage; -extern Instrumentation *InstrAlloc(int instrument_options); +/* + * The top instrumentation represents a running total of the current backend + * WAL/buffer usage information. This will not be updated immediately, but + * rather when the current stack entry gets accumulated which typically happens + * at query end. + * + * Care must be taken when utilizing this in the parallel worker context: + * Parallel workers will report back their instrumentation to the caller, + * and this gets added to the caller's stack. If this were to be used in the + * shared memory stats infrastructure it would need to be skipped on parallel + * workers to avoid double counting. + */ +extern PGDLLIMPORT Instrumentation instr_top; + +/* + * The instrumentation stack state. The 'current' field points to the + * currently active stack entry that is getting updated as activity happens, + * and will be accumulated to parent stacks when it gets finalized by + * InstrStop (for non-executor use cases), ExecFinalizeNodeInstrumentation + * (executor finish) or ResOwnerReleaseInstrumentation on abort. + */ +extern PGDLLIMPORT InstrStackState instr_stack; + +extern void InstrStackGrow(void); + +/* + * Pushes the stack so that all WAL/buffer usage updates go to the passed in + * instrumentation entry. + * + * See note on InstrPopStack regarding safe use of these functions. + */ +static inline void +InstrPushStack(Instrumentation *instr) +{ + if (unlikely(instr_stack.stack_size == instr_stack.stack_space)) + InstrStackGrow(); + + instr_stack.entries[instr_stack.stack_size++] = instr; + instr_stack.current = instr; +} + +/* + * Pops the stack entry back to the previous one that was effective at + * InstrPushStack. + * + * Callers must ensure that no intermediate stack entries are skipped, to + * handle aborts correctly. If you're thinking of calling this in a PG_FINALLY + * block, consider instead using InstrStart + InstrStopFinalize which can skip + * intermediate stack entries. + */ +static inline void +InstrPopStack(Instrumentation *instr) +{ + Assert(instr_stack.stack_size > 0); + Assert(instr_stack.entries[instr_stack.stack_size - 1] == instr); + instr_stack.stack_size--; + instr_stack.current = instr_stack.stack_size > 0 + ? instr_stack.entries[instr_stack.stack_size - 1] + : &instr_top; +} + extern void InstrInitOptions(Instrumentation *instr, int instrument_options); extern void InstrStart(Instrumentation *instr); extern void InstrStop(Instrumentation *instr); +extern void InstrStopFinalize(Instrumentation *instr); +extern void InstrFinalizeChild(Instrumentation *instr, Instrumentation *parent); +extern void InstrAccumStack(Instrumentation *dst, Instrumentation *add); -extern NodeInstrumentation *InstrAllocNode(int instrument_options, - bool async_mode); +extern QueryInstrumentation *InstrQueryAlloc(int instrument_options); +extern void InstrQueryStart(QueryInstrumentation *instr); +extern void InstrQueryStop(QueryInstrumentation *instr); +extern void InstrQueryStopFinalize(QueryInstrumentation *instr); +extern void InstrQueryRememberChild(QueryInstrumentation *parent, Instrumentation *instr); + +pg_nodiscard extern QueryInstrumentation *InstrStartParallelQuery(void); +extern void InstrEndParallelQuery(QueryInstrumentation *qinstr, BufferUsage *bufusage, WalUsage *walusage); +extern void InstrAccumParallelQuery(BufferUsage *bufusage, WalUsage *walusage); + +extern NodeInstrumentation *InstrAllocNode(QueryInstrumentation *qinstr, bool async_mode); extern void InstrInitNode(NodeInstrumentation *instr, int instrument_options); extern void InstrStartNode(NodeInstrumentation *instr); extern void InstrStopNode(NodeInstrumentation *instr, double nTuples); @@ -141,35 +294,36 @@ extern void InstrUpdateTupleCount(NodeInstrumentation *instr, double nTuples); extern void InstrEndLoop(NodeInstrumentation *instr); extern void InstrAggNode(NodeInstrumentation *dst, NodeInstrumentation *add); -extern TriggerInstrumentation *InstrAllocTrigger(int n, int instrument_options); -extern void InstrStartTrigger(TriggerInstrumentation *tginstr); +extern TriggerInstrumentation *InstrAllocTrigger(QueryInstrumentation *qinstr, int n); +extern void InstrStartTrigger(QueryInstrumentation *qinstr, + TriggerInstrumentation *tginstr); extern void InstrStopTrigger(TriggerInstrumentation *tginstr, int firings); -extern void InstrStartParallelQuery(void); -extern void InstrEndParallelQuery(BufferUsage *bufusage, WalUsage *walusage); -extern void InstrAccumParallelQuery(BufferUsage *bufusage, WalUsage *walusage); -extern void BufferUsageAccumDiff(BufferUsage *dst, - const BufferUsage *add, const BufferUsage *sub); +extern void BufferUsageAdd(BufferUsage *dst, const BufferUsage *add); +extern void WalUsageAdd(WalUsage *dst, const WalUsage *add); extern void WalUsageAccumDiff(WalUsage *dst, const WalUsage *add, const WalUsage *sub); #define INSTR_BUFUSAGE_INCR(fld) do { \ - pgBufferUsage.fld++; \ + instr_stack.current->bufusage.fld++; \ } while(0) #define INSTR_BUFUSAGE_ADD(fld,val) do { \ - pgBufferUsage.fld += (val); \ + instr_stack.current->bufusage.fld += (val); \ } while(0) #define INSTR_BUFUSAGE_TIME_ADD(fld,val) do { \ - INSTR_TIME_ADD(pgBufferUsage.fld, val); \ + INSTR_TIME_ADD(instr_stack.current->bufusage.fld, val); \ } while (0) #define INSTR_BUFUSAGE_TIME_ACCUM_DIFF(fld,endval,startval) do { \ - INSTR_TIME_ACCUM_DIFF(pgBufferUsage.fld, endval, startval); \ + INSTR_TIME_ACCUM_DIFF(instr_stack.current->bufusage.fld, endval, startval); \ } while (0) + #define INSTR_WALUSAGE_INCR(fld) do { \ pgWalUsage.fld++; \ + instr_stack.current->walusage.fld++; \ } while(0) #define INSTR_WALUSAGE_ADD(fld,val) do { \ pgWalUsage.fld += (val); \ + instr_stack.current->walusage.fld += (val); \ } while(0) #endif /* INSTRUMENT_H */ diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h index 502ad4f2da5..aef1003f608 100644 --- a/src/include/nodes/execnodes.h +++ b/src/include/nodes/execnodes.h @@ -53,6 +53,7 @@ typedef struct Instrumentation Instrumentation; typedef struct pairingheap pairingheap; typedef struct PlanState PlanState; typedef struct QueryEnvironment QueryEnvironment; +typedef struct QueryInstrumentation QueryInstrumentation; typedef struct RelationData *Relation; typedef Relation *RelationPtr; typedef struct ScanKeyData ScanKeyData; @@ -731,7 +732,7 @@ typedef struct EState * ExecutorRun() calls. */ int es_top_eflags; /* eflags passed to ExecutorStart */ - int es_instrument; /* OR of InstrumentOption flags */ + QueryInstrumentation *es_instrument; /* query-level instrumentation */ bool es_finished; /* true when ExecutorFinish is done */ List *es_exprcontexts; /* List of ExprContexts within EState */ diff --git a/src/include/utils/resowner.h b/src/include/utils/resowner.h index eb6033b4fdb..5463bc921f0 100644 --- a/src/include/utils/resowner.h +++ b/src/include/utils/resowner.h @@ -75,6 +75,7 @@ typedef uint32 ResourceReleasePriority; #define RELEASE_PRIO_SNAPSHOT_REFS 500 #define RELEASE_PRIO_FILES 600 #define RELEASE_PRIO_WAITEVENTSETS 700 +#define RELEASE_PRIO_INSTRUMENTATION 800 /* 0 is considered invalid */ #define RELEASE_PRIO_FIRST 1 diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list index 00dd1fc6ff9..d3203ac5e9a 100644 --- a/src/tools/pgindent/typedefs.list +++ b/src/tools/pgindent/typedefs.list @@ -1341,6 +1341,7 @@ InjectionPointSharedState InjectionPointsCtl InlineCodeBlock InsertStmt +InstrStackState Instrumentation Int128AggState Int8TransTypeData @@ -2463,6 +2464,7 @@ QueryCompletion QueryDesc QueryEnvironment QueryInfo +QueryInstrumentation QueryItem QueryItemType QueryMode -- 2.47.1