Hi Michael,
Le jeu. 3 oct. 2024 à 09:15, Michael Paquier <mich...@paquier.xyz> a écrit :
> On Thu, Aug 29, 2024 at 10:08:23PM +0200, Guillaume Lelarge wrote:
> > This patch was a bit discussed on [1], and with more details on [2]. It's
> > based on another patch sent in 2022 (see [3]). It introduces seven new
> > columns in pg_stat_statements:
> >
> > * parallelized_queries_planned, number of times the query has been
> planned
> > to be parallelized,
> > * parallelized_queries_launched, number of times the query has been
> > executed with parallelization,
>
> Comparing the numbers of workers planned and launched with the number
> of times a query has been called and planned should provide a rather
> good equivalent, no? I am not sure that these two are mandatory to
> have.
>
>
I'm not sure I follow. That would mean that every time a query is executed,
it always gets the same amount of workers. Which is not guaranteed to be
true.
I would agree, though, that parallelized_queries_launched is probably not
that interesting. I could get rid of it if you think it should go away.
> * parallelized_workers_planned, number of parallel workers planned for
> > this query,
> > * parallelized_workers_launched, number of parallel workers executed for
> > this query,
>
> Yep. Definitely OK with these two. There is an overlap with what
> Benoit has sent here when it comes to publish this data to the
> executor state:
>
> https://www.postgresql.org/message-id/783bc7f7-659a-42fa-99dd-ee0565644...@dalibo.com
>
>
Well, I don't see this as an overlap. Rather more information.
> > * parallelized_nodes, number of parallelized nodes,
> > * parallelized_nodes_all_workers, number of parallelized nodes which had
> > all requested workers,
> >
> > * parallelized_nodes_no_worker, number of parallelized nodes which had
> no
> > requested workers.
>
> I can see why you want to register this extra data on a node-basis,
> but how does that help when it comes to tune the parallel GUCs? We
> cannot control them at node level and the launched/planned ratio
> offers an equivalent of that. Not exactly, but that's enough to get a
> picture if there is a draught.
>
>
On this, I would agree with you. They are not that particularly useful to
get better setting for parallel GUCs. I can drop them if you want.
> A test script (test2.sql) is attached. You can execute it with "psql -Xef
> > test2.sql your_database" (your_database should not contain a t1 table as
> it
> > will be dropped and recreated).
>
> Let's add proper regression tests instead, including
> oldextversions.sql as this bumps the version of the module. See for
> example the tests of 6fd5071909a2 that can force workers to spawn
> for BRIN and btree queries, validating some of the stats published
> here.
>
Did this on the v2 version of the patch (attached here).
Thanks for your review. If you want the parallelized_queries_launched
column and the parallelized_nodes_* columns dropped, I can do that on a v3
patch.
Regards.
--
Guillaume.
From 8228f8d521eb9853cba6b5d185e4d9965f29d398 Mon Sep 17 00:00:00 2001
From: Guillaume Lelarge <guillaume.lela...@dalibo.com>
Date: Wed, 28 Aug 2024 15:30:05 +0200
Subject: [PATCH v2] Add parallel columns to pg_stat_statements
There are seven new columns:
* parallelized_queries_planned (number of times the query has been planned to
be parallelized),
* parallel_ized_querieslaunched (number of times the query has been executed
with parallelization),
* parallelized_workers_planned (number of parallel workers planned for
this query),
* parallelized_workers_launched (number of parallel workers executed for
this query),
* parallelized_nodes (number of parallelized nodes),
* parallelized_nodes_all_workers (number of parallelized nodes which
had all requested workers),
* parallelized_nodes_no_worker (number of parallelized nodes which had
no requested workers).
These new columns will help to monitor and better configure query
parallelization.
---
contrib/pg_stat_statements/Makefile | 4 +-
.../expected/oldextversions.out | 69 +++++++++++
.../pg_stat_statements/expected/parallel.out | 38 ++++++
.../pg_stat_statements--1.11--1.12.sql | 80 +++++++++++++
.../pg_stat_statements/pg_stat_statements.c | 108 ++++++++++++++++--
.../pg_stat_statements.control | 2 +-
.../pg_stat_statements/sql/oldextversions.sql | 5 +
contrib/pg_stat_statements/sql/parallel.sql | 27 +++++
doc/src/sgml/pgstatstatements.sgml | 63 ++++++++++
src/backend/executor/execUtils.c | 7 ++
src/backend/executor/nodeGather.c | 9 +-
src/backend/executor/nodeGatherMerge.c | 8 ++
src/include/nodes/execnodes.h | 6 +
13 files changed, 415 insertions(+), 11 deletions(-)
create mode 100644 contrib/pg_stat_statements/expected/parallel.out
create mode 100644 contrib/pg_stat_statements/pg_stat_statements--1.11--1.12.sql
create mode 100644 contrib/pg_stat_statements/sql/parallel.sql
diff --git a/contrib/pg_stat_statements/Makefile b/contrib/pg_stat_statements/Makefile
index 1622b43ded..0caeea0c03 100644
--- a/contrib/pg_stat_statements/Makefile
+++ b/contrib/pg_stat_statements/Makefile
@@ -7,7 +7,7 @@ OBJS = \
EXTENSION = pg_stat_statements
DATA = pg_stat_statements--1.4.sql \
- pg_stat_statements--1.10--1.11.sql \
+ pg_stat_statements--1.11--1.12.sql pg_stat_statements--1.10--1.11.sql \
pg_stat_statements--1.9--1.10.sql pg_stat_statements--1.8--1.9.sql \
pg_stat_statements--1.7--1.8.sql pg_stat_statements--1.6--1.7.sql \
pg_stat_statements--1.5--1.6.sql pg_stat_statements--1.4--1.5.sql \
@@ -19,7 +19,7 @@ LDFLAGS_SL += $(filter -lm, $(LIBS))
REGRESS_OPTS = --temp-config $(top_srcdir)/contrib/pg_stat_statements/pg_stat_statements.conf
REGRESS = select dml cursors utility level_tracking planning \
- user_activity wal entry_timestamp privileges extended cleanup \
+ user_activity wal entry_timestamp privileges extended parallel cleanup \
oldextversions
# Disabled because these tests require "shared_preload_libraries=pg_stat_statements",
# which typical installcheck users do not have (e.g. buildfarm clients).
diff --git a/contrib/pg_stat_statements/expected/oldextversions.out b/contrib/pg_stat_statements/expected/oldextversions.out
index 5842c930e5..a4e8dc0bde 100644
--- a/contrib/pg_stat_statements/expected/oldextversions.out
+++ b/contrib/pg_stat_statements/expected/oldextversions.out
@@ -342,4 +342,73 @@ SELECT pg_stat_statements_reset() IS NOT NULL AS t;
t
(1 row)
+-- New functions and views for pg_stat_statements in 1.12
+AlTER EXTENSION pg_stat_statements UPDATE TO '1.12';
+\d pg_stat_statements
+ View "public.pg_stat_statements"
+ Column | Type | Collation | Nullable | Default
+--------------------------------+--------------------------+-----------+----------+---------
+ userid | oid | | |
+ dbid | oid | | |
+ toplevel | boolean | | |
+ queryid | bigint | | |
+ query | text | | |
+ plans | bigint | | |
+ total_plan_time | double precision | | |
+ min_plan_time | double precision | | |
+ max_plan_time | double precision | | |
+ mean_plan_time | double precision | | |
+ stddev_plan_time | double precision | | |
+ calls | bigint | | |
+ total_exec_time | double precision | | |
+ min_exec_time | double precision | | |
+ max_exec_time | double precision | | |
+ mean_exec_time | double precision | | |
+ stddev_exec_time | double precision | | |
+ rows | bigint | | |
+ shared_blks_hit | bigint | | |
+ shared_blks_read | bigint | | |
+ shared_blks_dirtied | bigint | | |
+ shared_blks_written | bigint | | |
+ local_blks_hit | bigint | | |
+ local_blks_read | bigint | | |
+ local_blks_dirtied | bigint | | |
+ local_blks_written | bigint | | |
+ temp_blks_read | bigint | | |
+ temp_blks_written | bigint | | |
+ shared_blk_read_time | double precision | | |
+ shared_blk_write_time | double precision | | |
+ local_blk_read_time | double precision | | |
+ local_blk_write_time | double precision | | |
+ temp_blk_read_time | double precision | | |
+ temp_blk_write_time | double precision | | |
+ wal_records | bigint | | |
+ wal_fpi | bigint | | |
+ wal_bytes | numeric | | |
+ jit_functions | bigint | | |
+ jit_generation_time | double precision | | |
+ jit_inlining_count | bigint | | |
+ jit_inlining_time | double precision | | |
+ jit_optimization_count | bigint | | |
+ jit_optimization_time | double precision | | |
+ jit_emission_count | bigint | | |
+ jit_emission_time | double precision | | |
+ jit_deform_count | bigint | | |
+ jit_deform_time | double precision | | |
+ parallelized_queries_planned | bigint | | |
+ parallelized_queries_launched | bigint | | |
+ parallelized_workers_planned | bigint | | |
+ parallelized_workers_launched | bigint | | |
+ parallelized_nodes | bigint | | |
+ parallelized_nodes_all_workers | bigint | | |
+ parallelized_nodes_no_worker | bigint | | |
+ stats_since | timestamp with time zone | | |
+ minmax_stats_since | timestamp with time zone | | |
+
+SELECT count(*) > 0 AS has_data FROM pg_stat_statements;
+ has_data
+----------
+ t
+(1 row)
+
DROP EXTENSION pg_stat_statements;
diff --git a/contrib/pg_stat_statements/expected/parallel.out b/contrib/pg_stat_statements/expected/parallel.out
new file mode 100644
index 0000000000..1379d10e97
--- /dev/null
+++ b/contrib/pg_stat_statements/expected/parallel.out
@@ -0,0 +1,38 @@
+--
+-- Validate parallelization generation metrics
+--
+SET parallel_setup_cost TO 0;
+SET min_parallel_table_scan_size TO '100kB';
+SET pg_stat_statements.track_utility = FALSE;
+CREATE TABLE pgss_parallel_tab (a int);
+INSERT INTO pgss_parallel_tab SELECT generate_series(1, 10_000);
+SELECT count(*) FROM pgss_parallel_tab;
+ count
+-------
+ 10000
+(1 row)
+
+DROP TABLE pgss_parallel_tab;
+-- Check parallelization metrics are generated for the SELECT statement
+SELECT query,
+ parallelized_queries_planned > 0 AS parallelized_queries_planned_generated,
+ parallelized_queries_launched > 0 AS parallelized_queries_launched_generated,
+ parallelized_workers_planned > 0 AS parallelized_workers_planned_generated,
+ parallelized_workers_launched > 0 AS parallelized_workers_launched_generated,
+ parallelized_nodes > 0 AS parallelized_nodes_generated,
+ parallelized_nodes_all_workers > 0 AS parallelized_nodes_all_workers_generated,
+ parallelized_nodes_no_worker = 0 AS parallelized_nodes_no_worker_generated
+FROM pg_stat_statements
+WHERE query LIKE 'SELECT count%'
+ORDER BY query COLLATE "C";
+ query | parallelized_queries_planned_generated | parallelized_queries_launched_generated | parallelized_workers_planned_generated | parallelized_workers_launched_generated | parallelized_nodes_generated | parallelized_nodes_all_workers_generated | parallelized_nodes_no_worker_generated
+----------------------------------------+----------------------------------------+-----------------------------------------+----------------------------------------+-----------------------------------------+------------------------------+------------------------------------------+----------------------------------------
+ SELECT count(*) FROM pgss_parallel_tab | t | t | t | t | t | t | t
+(1 row)
+
+SELECT pg_stat_statements_reset() IS NOT NULL AS t;
+ t
+---
+ t
+(1 row)
+
diff --git a/contrib/pg_stat_statements/pg_stat_statements--1.11--1.12.sql b/contrib/pg_stat_statements/pg_stat_statements--1.11--1.12.sql
new file mode 100644
index 0000000000..6f4fe8be48
--- /dev/null
+++ b/contrib/pg_stat_statements/pg_stat_statements--1.11--1.12.sql
@@ -0,0 +1,80 @@
+/* contrib/pg_stat_statements/pg_stat_statements--1.11--1.12.sql */
+
+-- complain if script is sourced in psql, rather than via ALTER EXTENSION
+\echo Use "ALTER EXTENSION pg_stat_statements UPDATE TO '1.12'" to load this file. \quit
+
+/* First we have to remove them from the extension */
+ALTER EXTENSION pg_stat_statements DROP VIEW pg_stat_statements;
+ALTER EXTENSION pg_stat_statements DROP FUNCTION pg_stat_statements(boolean);
+
+/* Then we can drop them */
+DROP VIEW pg_stat_statements;
+DROP FUNCTION pg_stat_statements(boolean);
+
+/* Now redefine */
+CREATE FUNCTION pg_stat_statements(IN showtext boolean,
+ OUT userid oid,
+ OUT dbid oid,
+ OUT toplevel bool,
+ OUT queryid bigint,
+ OUT query text,
+ OUT plans int8,
+ OUT total_plan_time float8,
+ OUT min_plan_time float8,
+ OUT max_plan_time float8,
+ OUT mean_plan_time float8,
+ OUT stddev_plan_time float8,
+ OUT calls int8,
+ OUT total_exec_time float8,
+ OUT min_exec_time float8,
+ OUT max_exec_time float8,
+ OUT mean_exec_time float8,
+ OUT stddev_exec_time float8,
+ OUT rows int8,
+ OUT shared_blks_hit int8,
+ OUT shared_blks_read int8,
+ OUT shared_blks_dirtied int8,
+ OUT shared_blks_written int8,
+ OUT local_blks_hit int8,
+ OUT local_blks_read int8,
+ OUT local_blks_dirtied int8,
+ OUT local_blks_written int8,
+ OUT temp_blks_read int8,
+ OUT temp_blks_written int8,
+ OUT shared_blk_read_time float8,
+ OUT shared_blk_write_time float8,
+ OUT local_blk_read_time float8,
+ OUT local_blk_write_time float8,
+ OUT temp_blk_read_time float8,
+ OUT temp_blk_write_time float8,
+ OUT wal_records int8,
+ OUT wal_fpi int8,
+ OUT wal_bytes numeric,
+ OUT jit_functions int8,
+ OUT jit_generation_time float8,
+ OUT jit_inlining_count int8,
+ OUT jit_inlining_time float8,
+ OUT jit_optimization_count int8,
+ OUT jit_optimization_time float8,
+ OUT jit_emission_count int8,
+ OUT jit_emission_time float8,
+ OUT jit_deform_count int8,
+ OUT jit_deform_time float8,
+ OUT parallelized_queries_planned int8,
+ OUT parallelized_queries_launched int8,
+ OUT parallelized_workers_planned int8,
+ OUT parallelized_workers_launched int8,
+ OUT parallelized_nodes int8,
+ OUT parallelized_nodes_all_workers int8,
+ OUT parallelized_nodes_no_worker int8,
+ OUT stats_since timestamp with time zone,
+ OUT minmax_stats_since timestamp with time zone
+)
+RETURNS SETOF record
+AS 'MODULE_PATHNAME', 'pg_stat_statements_1_12'
+LANGUAGE C STRICT VOLATILE PARALLEL SAFE;
+
+CREATE VIEW pg_stat_statements AS
+ SELECT * FROM pg_stat_statements(true);
+
+GRANT SELECT ON pg_stat_statements TO PUBLIC;
diff --git a/contrib/pg_stat_statements/pg_stat_statements.c b/contrib/pg_stat_statements/pg_stat_statements.c
index 5765ef49b4..da94d64011 100644
--- a/contrib/pg_stat_statements/pg_stat_statements.c
+++ b/contrib/pg_stat_statements/pg_stat_statements.c
@@ -113,6 +113,7 @@ typedef enum pgssVersion
PGSS_V1_9,
PGSS_V1_10,
PGSS_V1_11,
+ PGSS_V1_12,
} pgssVersion;
typedef enum pgssStoreKind
@@ -204,6 +205,13 @@ typedef struct Counters
int64 jit_emission_count; /* number of times emission time has been
* > 0 */
double jit_emission_time; /* total time to emit jit code */
+ int64 parallelized_queries_planned; /* # of times query was planned to use parallelism */
+ int64 parallelized_queries_launched; /* # of times query was executed using parallelism */
+ int64 parallelized_workers_planned; /* # of parallel workers planned */
+ int64 parallelized_workers_launched; /* # of parallel workers launched */
+ int64 parallelized_nodes; /* # of parallelized nodes */
+ int64 parallelized_nodes_all_workers; /* # of parallelized nodes with all workers */
+ int64 parallelized_nodes_no_worker; /* # of parallelized nodes with no workers */
} Counters;
/*
@@ -317,6 +325,7 @@ PG_FUNCTION_INFO_V1(pg_stat_statements_1_8);
PG_FUNCTION_INFO_V1(pg_stat_statements_1_9);
PG_FUNCTION_INFO_V1(pg_stat_statements_1_10);
PG_FUNCTION_INFO_V1(pg_stat_statements_1_11);
+PG_FUNCTION_INFO_V1(pg_stat_statements_1_12);
PG_FUNCTION_INFO_V1(pg_stat_statements);
PG_FUNCTION_INFO_V1(pg_stat_statements_info);
@@ -347,7 +356,14 @@ static void pgss_store(const char *query, uint64 queryId,
const BufferUsage *bufusage,
const WalUsage *walusage,
const struct JitInstrumentation *jitusage,
- JumbleState *jstate);
+ JumbleState *jstate,
+ bool parallelized_queries_planned,
+ bool parallelized_queries_launched,
+ int parallelized_workers_planned,
+ int parallelized_workers_launched,
+ int parallelized_nodes,
+ int parallelized_nodes_all_workers,
+ int parallelized_nodes_no_worker);
static void pg_stat_statements_internal(FunctionCallInfo fcinfo,
pgssVersion api_version,
bool showtext);
@@ -867,7 +883,14 @@ pgss_post_parse_analyze(ParseState *pstate, Query *query, JumbleState *jstate)
NULL,
NULL,
NULL,
- jstate);
+ jstate,
+ false,
+ false,
+ 0,
+ 0,
+ 0,
+ 0,
+ 0);
}
/*
@@ -945,7 +968,14 @@ pgss_planner(Query *parse,
&bufusage,
&walusage,
NULL,
- NULL);
+ NULL,
+ false,
+ false,
+ 0,
+ 0,
+ 0,
+ 0,
+ 0);
}
else
{
@@ -1078,7 +1108,14 @@ pgss_ExecutorEnd(QueryDesc *queryDesc)
&queryDesc->totaltime->bufusage,
&queryDesc->totaltime->walusage,
queryDesc->estate->es_jit ? &queryDesc->estate->es_jit->instr : NULL,
- NULL);
+ NULL,
+ queryDesc->plannedstmt->parallelModeNeeded,
+ queryDesc->estate->es_used_parallel_mode,
+ queryDesc->estate->es_parallelized_workers_planned,
+ queryDesc->estate->es_parallelized_workers_launched,
+ queryDesc->estate->es_parallelized_nodes,
+ queryDesc->estate->es_parallelized_nodes_all_workers,
+ queryDesc->estate->es_parallelized_nodes_no_worker);
}
if (prev_ExecutorEnd)
@@ -1209,7 +1246,14 @@ pgss_ProcessUtility(PlannedStmt *pstmt, const char *queryString,
&bufusage,
&walusage,
NULL,
- NULL);
+ NULL,
+ false,
+ false,
+ 0,
+ 0,
+ 0,
+ 0,
+ 0);
}
else
{
@@ -1270,7 +1314,14 @@ pgss_store(const char *query, uint64 queryId,
const BufferUsage *bufusage,
const WalUsage *walusage,
const struct JitInstrumentation *jitusage,
- JumbleState *jstate)
+ JumbleState *jstate,
+ bool parallelized_queries_planned,
+ bool parallelized_queries_launched,
+ int parallelized_workers_planned,
+ int parallelized_workers_launched,
+ int parallelized_nodes,
+ int parallelized_nodes_all_workers,
+ int parallelized_nodes_no_worker)
{
pgssHashKey key;
pgssEntry *entry;
@@ -1472,6 +1523,20 @@ pgss_store(const char *query, uint64 queryId,
entry->counters.jit_emission_count++;
entry->counters.jit_emission_time += INSTR_TIME_GET_MILLISEC(jitusage->emission_counter);
}
+ /* inc parallel counters */
+ if (parallelized_queries_planned)
+ {
+ entry->counters.parallelized_queries_planned += 1;
+ }
+ if (parallelized_queries_launched)
+ {
+ entry->counters.parallelized_queries_launched += 1;
+ }
+ entry->counters.parallelized_workers_planned += parallelized_workers_planned;
+ entry->counters.parallelized_workers_launched += parallelized_workers_launched;
+ entry->counters.parallelized_nodes += parallelized_nodes;
+ entry->counters.parallelized_nodes_all_workers += parallelized_nodes_all_workers;
+ entry->counters.parallelized_nodes_no_worker += parallelized_nodes_no_worker;
SpinLockRelease(&entry->mutex);
}
@@ -1539,7 +1604,8 @@ pg_stat_statements_reset(PG_FUNCTION_ARGS)
#define PG_STAT_STATEMENTS_COLS_V1_9 33
#define PG_STAT_STATEMENTS_COLS_V1_10 43
#define PG_STAT_STATEMENTS_COLS_V1_11 49
-#define PG_STAT_STATEMENTS_COLS 49 /* maximum of above */
+#define PG_STAT_STATEMENTS_COLS_V1_12 56
+#define PG_STAT_STATEMENTS_COLS 56 /* maximum of above */
/*
* Retrieve statement statistics.
@@ -1551,6 +1617,16 @@ pg_stat_statements_reset(PG_FUNCTION_ARGS)
* expected API version is identified by embedding it in the C name of the
* function. Unfortunately we weren't bright enough to do that for 1.1.
*/
+Datum
+pg_stat_statements_1_12(PG_FUNCTION_ARGS)
+{
+ bool showtext = PG_GETARG_BOOL(0);
+
+ pg_stat_statements_internal(fcinfo, PGSS_V1_12, showtext);
+
+ return (Datum) 0;
+}
+
Datum
pg_stat_statements_1_11(PG_FUNCTION_ARGS)
{
@@ -1695,6 +1771,10 @@ pg_stat_statements_internal(FunctionCallInfo fcinfo,
if (api_version != PGSS_V1_11)
elog(ERROR, "incorrect number of output arguments");
break;
+ case PG_STAT_STATEMENTS_COLS_V1_12:
+ if (api_version != PGSS_V1_12)
+ elog(ERROR, "incorrect number of output arguments");
+ break;
default:
elog(ERROR, "incorrect number of output arguments");
}
@@ -1932,6 +2012,19 @@ pg_stat_statements_internal(FunctionCallInfo fcinfo,
{
values[i++] = Int64GetDatumFast(tmp.jit_deform_count);
values[i++] = Float8GetDatumFast(tmp.jit_deform_time);
+ }
+ if (api_version >= PGSS_V1_12)
+ {
+ values[i++] = Int64GetDatumFast(tmp.parallelized_queries_planned);
+ values[i++] = Int64GetDatumFast(tmp.parallelized_queries_launched);
+ values[i++] = Int64GetDatumFast(tmp.parallelized_workers_planned);
+ values[i++] = Int64GetDatumFast(tmp.parallelized_workers_launched);
+ values[i++] = Int64GetDatumFast(tmp.parallelized_nodes);
+ values[i++] = Int64GetDatumFast(tmp.parallelized_nodes_all_workers);
+ values[i++] = Int64GetDatumFast(tmp.parallelized_nodes_no_worker);
+ }
+ if (api_version >= PGSS_V1_11)
+ {
values[i++] = TimestampTzGetDatum(stats_since);
values[i++] = TimestampTzGetDatum(minmax_stats_since);
}
@@ -1944,6 +2037,7 @@ pg_stat_statements_internal(FunctionCallInfo fcinfo,
api_version == PGSS_V1_9 ? PG_STAT_STATEMENTS_COLS_V1_9 :
api_version == PGSS_V1_10 ? PG_STAT_STATEMENTS_COLS_V1_10 :
api_version == PGSS_V1_11 ? PG_STAT_STATEMENTS_COLS_V1_11 :
+ api_version == PGSS_V1_12 ? PG_STAT_STATEMENTS_COLS_V1_12 :
-1 /* fail if you forget to update this assert */ ));
tuplestore_putvalues(rsinfo->setResult, rsinfo->setDesc, values, nulls);
diff --git a/contrib/pg_stat_statements/pg_stat_statements.control b/contrib/pg_stat_statements/pg_stat_statements.control
index 8a76106ec6..d45ebc12e3 100644
--- a/contrib/pg_stat_statements/pg_stat_statements.control
+++ b/contrib/pg_stat_statements/pg_stat_statements.control
@@ -1,5 +1,5 @@
# pg_stat_statements extension
comment = 'track planning and execution statistics of all SQL statements executed'
-default_version = '1.11'
+default_version = '1.12'
module_pathname = '$libdir/pg_stat_statements'
relocatable = true
diff --git a/contrib/pg_stat_statements/sql/oldextversions.sql b/contrib/pg_stat_statements/sql/oldextversions.sql
index 38d5505d0d..13b8ca2858 100644
--- a/contrib/pg_stat_statements/sql/oldextversions.sql
+++ b/contrib/pg_stat_statements/sql/oldextversions.sql
@@ -58,4 +58,9 @@ SELECT count(*) > 0 AS has_data FROM pg_stat_statements;
SELECT pg_get_functiondef('pg_stat_statements_reset'::regproc);
SELECT pg_stat_statements_reset() IS NOT NULL AS t;
+-- New functions and views for pg_stat_statements in 1.12
+AlTER EXTENSION pg_stat_statements UPDATE TO '1.12';
+\d pg_stat_statements
+SELECT count(*) > 0 AS has_data FROM pg_stat_statements;
+
DROP EXTENSION pg_stat_statements;
diff --git a/contrib/pg_stat_statements/sql/parallel.sql b/contrib/pg_stat_statements/sql/parallel.sql
new file mode 100644
index 0000000000..bdc9be6e1e
--- /dev/null
+++ b/contrib/pg_stat_statements/sql/parallel.sql
@@ -0,0 +1,27 @@
+--
+-- Validate parallelization generation metrics
+--
+
+SET parallel_setup_cost TO 0;
+SET min_parallel_table_scan_size TO '100kB';
+SET pg_stat_statements.track_utility = FALSE;
+
+CREATE TABLE pgss_parallel_tab (a int);
+INSERT INTO pgss_parallel_tab SELECT generate_series(1, 10_000);
+SELECT count(*) FROM pgss_parallel_tab;
+DROP TABLE pgss_parallel_tab;
+
+-- Check parallelization metrics are generated for the SELECT statement
+SELECT query,
+ parallelized_queries_planned > 0 AS parallelized_queries_planned_generated,
+ parallelized_queries_launched > 0 AS parallelized_queries_launched_generated,
+ parallelized_workers_planned > 0 AS parallelized_workers_planned_generated,
+ parallelized_workers_launched > 0 AS parallelized_workers_launched_generated,
+ parallelized_nodes > 0 AS parallelized_nodes_generated,
+ parallelized_nodes_all_workers > 0 AS parallelized_nodes_all_workers_generated,
+ parallelized_nodes_no_worker = 0 AS parallelized_nodes_no_worker_generated
+FROM pg_stat_statements
+WHERE query LIKE 'SELECT count%'
+ORDER BY query COLLATE "C";
+
+SELECT pg_stat_statements_reset() IS NOT NULL AS t;
diff --git a/doc/src/sgml/pgstatstatements.sgml b/doc/src/sgml/pgstatstatements.sgml
index 9b0aff73b1..3fe84ba994 100644
--- a/doc/src/sgml/pgstatstatements.sgml
+++ b/doc/src/sgml/pgstatstatements.sgml
@@ -527,6 +527,69 @@
</para></entry>
</row>
+ <row>
+ <entry role="catalog_table_entry"><para role="column_definition">
+ <structfield>parallelized_queries_planned</structfield> <type>bigint</type>
+ </para>
+ <para>
+ Number of times the statement was planned to use parallelism.
+ </para></entry>
+ </row>
+
+ <row>
+ <entry role="catalog_table_entry"><para role="column_definition">
+ <structfield>parallel_queries_launched</structfield> <type>bigint</type>
+ </para>
+ <para>
+ Number of times that the statement was executed using parallelism
+ </para></entry>
+ </row>
+
+ <row>
+ <entry role="catalog_table_entry"><para role="column_definition">
+ <structfield>parallelized_workers_planned</structfield> <type>bigint</type>
+ </para>
+ <para>
+ Number of parallel workers planned for the query.
+ </para></entry>
+ </row>
+
+ <row>
+ <entry role="catalog_table_entry"><para role="column_definition">
+ <structfield>parallel_workers_launched</structfield> <type>bigint</type>
+ </para>
+ <para>
+ Number of parallel workers executed for the query.
+ </para></entry>
+ </row>
+
+ <row>
+ <entry role="catalog_table_entry"><para role="column_definition">
+ <structfield>parallelized_nodes</structfield> <type>bigint</type>
+ </para>
+ <para>
+ Number of parallelized nodes
+ </para></entry>
+ </row>
+
+ <row>
+ <entry role="catalog_table_entry"><para role="column_definition">
+ <structfield>parallelized_nodes_all_workers</structfield> <type>bigint</type>
+ </para>
+ <para>
+ Number of parallelized nodes that got all workers
+ </para></entry>
+ </row>
+
+ <row>
+ <entry role="catalog_table_entry"><para role="column_definition">
+ <structfield>parallelized_nodes_no_worker</structfield> <type>bigint</type>
+ </para>
+ <para>
+ Number of parallelized nodes that got no worker at all
+ </para></entry>
+ </row>
+
<row>
<entry role="catalog_table_entry"><para role="column_definition">
<structfield>stats_since</structfield> <type>timestamp with time zone</type>
diff --git a/src/backend/executor/execUtils.c b/src/backend/executor/execUtils.c
index 5737f9f4eb..3b0715d604 100644
--- a/src/backend/executor/execUtils.c
+++ b/src/backend/executor/execUtils.c
@@ -158,10 +158,17 @@ CreateExecutorState(void)
estate->es_sourceText = NULL;
estate->es_use_parallel_mode = false;
+ estate->es_used_parallel_mode = false;
estate->es_jit_flags = 0;
estate->es_jit = NULL;
+ estate->es_parallelized_nodes = 0;
+ estate->es_parallelized_nodes_all_workers = 0;
+ estate->es_parallelized_nodes_no_worker = 0;
+ estate->es_parallelized_workers_launched = 0;
+ estate->es_parallelized_workers_planned = 0;
+
/*
* Return the executor state structure
*/
diff --git a/src/backend/executor/nodeGather.c b/src/backend/executor/nodeGather.c
index 5d4ffe989c..f1f772ecb6 100644
--- a/src/backend/executor/nodeGather.c
+++ b/src/backend/executor/nodeGather.c
@@ -181,7 +181,13 @@ ExecGather(PlanState *pstate)
LaunchParallelWorkers(pcxt);
/* We save # workers launched for the benefit of EXPLAIN */
node->nworkers_launched = pcxt->nworkers_launched;
-
+ if(pcxt->nworkers_launched > 0)
+ estate->es_used_parallel_mode = true;
+ estate->es_parallelized_nodes += 1;
+ estate->es_parallelized_workers_launched += pcxt->nworkers_launched;
+ estate->es_parallelized_workers_planned += pcxt->nworkers_to_launch;
+ if (pcxt->nworkers_to_launch == pcxt->nworkers_launched)
+ estate->es_parallelized_nodes_all_workers += 1;
/* Set up tuple queue readers to read the results. */
if (pcxt->nworkers_launched > 0)
{
@@ -198,6 +204,7 @@ ExecGather(PlanState *pstate)
/* No workers? Then never mind. */
node->nreaders = 0;
node->reader = NULL;
+ estate->es_parallelized_nodes_no_worker += 1;
}
node->nextreader = 0;
}
diff --git a/src/backend/executor/nodeGatherMerge.c b/src/backend/executor/nodeGatherMerge.c
index 45f6017c29..5678b25a86 100644
--- a/src/backend/executor/nodeGatherMerge.c
+++ b/src/backend/executor/nodeGatherMerge.c
@@ -222,6 +222,13 @@ ExecGatherMerge(PlanState *pstate)
LaunchParallelWorkers(pcxt);
/* We save # workers launched for the benefit of EXPLAIN */
node->nworkers_launched = pcxt->nworkers_launched;
+ if(pcxt->nworkers_launched > 0)
+ estate->es_used_parallel_mode = true;
+ estate->es_parallelized_nodes += 1;
+ estate->es_parallelized_workers_launched += pcxt->nworkers_launched;
+ estate->es_parallelized_workers_planned += pcxt->nworkers_to_launch;
+ if (pcxt->nworkers_to_launch == pcxt->nworkers_launched)
+ estate->es_parallelized_nodes_all_workers += 1;
/* Set up tuple queue readers to read the results. */
if (pcxt->nworkers_launched > 0)
@@ -239,6 +246,7 @@ ExecGatherMerge(PlanState *pstate)
/* No workers? Then never mind. */
node->nreaders = 0;
node->reader = NULL;
+ estate->es_parallelized_nodes_no_worker += 1;
}
}
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
index aab59d681c..d114e8d522 100644
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -707,6 +707,12 @@ typedef struct EState
struct EPQState *es_epq_active;
bool es_use_parallel_mode; /* can we use parallel workers? */
+ bool es_used_parallel_mode; /* was executed in parallel */
+ int es_parallelized_workers_launched;
+ int es_parallelized_workers_planned;
+ int es_parallelized_nodes; /* # of parallelized nodes */
+ int es_parallelized_nodes_all_workers; /* # of nodes with all workers launched */
+ int es_parallelized_nodes_no_worker; /* # of nodes with no workers launched */
/* The per-query shared memory area to use for parallel execution. */
struct dsa_area *es_query_dsa;
--
2.46.2
