agora inbox for pgsql-hackers@postgresql.org  
help / color / mirror / Atom feed
[PATCH v12 5/5] More refactoring
325+ messages / 2 participants
[nested] [flat]

* [PATCH v12 5/5] More refactoring
@ 2020-11-01 19:46 Justin Pryzby <pryzbyj@telsasoft.com>
  0 siblings, 0 replies; 325+ messages in thread

From: Justin Pryzby @ 2020-11-01 19:46 UTC (permalink / raw)

---
 src/backend/commands/indexcmds.c           | 183 +++++++++------------
 src/test/regress/expected/create_index.out |   4 +-
 2 files changed, 83 insertions(+), 104 deletions(-)

diff --git a/src/backend/commands/indexcmds.c b/src/backend/commands/indexcmds.c
index 08d44a1999..4586942960 100644
--- a/src/backend/commands/indexcmds.c
+++ b/src/backend/commands/indexcmds.c
@@ -102,8 +102,8 @@ static void ReindexMultipleInternal(List *relids,
 									ReindexParams *params);
 static bool ReindexRelationConcurrently(Oid relationOid,
 										ReindexParams *params);
-static List *ReindexIndexesConcurrently(List *indexIds, List *heapRelationIds,
-										int options, MemoryContext private_context);
+static List *ReindexIndexesConcurrently(List *indexIds, int options,
+										MemoryContext private_context);
 static void update_relispartition(Oid relationId, bool newval);
 static inline void set_indexsafe_procflags(void);
 
@@ -2625,8 +2625,8 @@ ReindexIndex(RangeVar *indexRelation, ReindexParams *params, bool isTopLevel)
 			.indexId = indOid,
 			/* other fields set later */
 		};
+
 		ReindexIndexesConcurrently(list_make1(&idxinfo),
-				list_make1_oid(IndexGetRelation(indOid, false)),
 				params->options, CurrentMemoryContext);
 	}
 	else
@@ -2961,12 +2961,11 @@ reindex_error_callback(void *arg)
 
 
 /*
- * Given a list of index oids, return a list of leaf partitions by removing
- * any intermediate parents.  heaprels is populated with the corresponding
- * tables.
+ * Given a list of index oids, return a new list of leaf partitions by
+ * excluding any intermediate parents.
  */
 static List *
-leaf_partitions(List *inhoids, int options, List **heaprels)
+leaf_partitions(List *inhoids, int options)
 {
 	List		*partitions = NIL;
 	ListCell	*lc;
@@ -2998,7 +2997,6 @@ leaf_partitions(List *inhoids, int options, List **heaprels)
 
 		/* Save partition OID in current MemoryContext */
 		partitions = lappend_oid(partitions, partoid);
-		*heaprels = lappend_oid(*heaprels, tableoid);
 	}
 
 	return partitions;
@@ -3010,13 +3008,11 @@ leaf_partitions(List *inhoids, int options, List **heaprels)
  *
  * Reindex a set of partitions, per the partitioned index or table given
  * by the caller.
- * XXX: should be further refactored with logic from ReindexRelationConcurrently
  */
 static void
 ReindexPartitions(Oid relid, ReindexParams *params, bool isTopLevel)
 {
-	List	   *partitions = NIL,
-			*heaprels = NIL;
+	List	   *partitions = NIL;
 	char		relkind = get_rel_relkind(relid);
 	char	   *relname = get_rel_name(relid);
 	char	   *relnamespace = get_namespace_name(get_rel_namespace(relid));
@@ -3070,7 +3066,7 @@ ReindexPartitions(Oid relid, ReindexParams *params, bool isTopLevel)
 	if (relkind == RELKIND_PARTITIONED_INDEX)
 	{
 		old_context = MemoryContextSwitchTo(reindex_context);
-		partitions = leaf_partitions(inhoids, params->options, &heaprels);
+		partitions = leaf_partitions(inhoids, params->options);
 		MemoryContextSwitchTo(old_context);
 	} else {
 		/* Loop over parent tables */
@@ -3083,7 +3079,7 @@ ReindexPartitions(Oid relid, ReindexParams *params, bool isTopLevel)
 			parttable = table_open(partoid, ShareLock);
 			old_context = MemoryContextSwitchTo(reindex_context);
 			partindexes = RelationGetIndexList(parttable);
-			partindexes = leaf_partitions(partindexes, params->options, &heaprels);
+			partindexes = leaf_partitions(partindexes, params->options);
 			partitions = list_concat(partitions, partindexes);
 
 			MemoryContextSwitchTo(old_context);
@@ -3092,10 +3088,9 @@ ReindexPartitions(Oid relid, ReindexParams *params, bool isTopLevel)
 	}
 
 	if ((params->options & REINDEXOPT_CONCURRENTLY) != 0 &&
-		relkind == RELKIND_PARTITIONED_INDEX &&
 		get_rel_persistence(relid) != RELPERSISTENCE_TEMP)
 	{
-		List			   *idxinfos = NIL;
+		List			    *idxinfos = NIL;
 		ReindexIndexInfo	*idxinfo;
 
 		old_context = MemoryContextSwitchTo(reindex_context);
@@ -3110,7 +3105,7 @@ ReindexPartitions(Oid relid, ReindexParams *params, bool isTopLevel)
 		MemoryContextSwitchTo(old_context);
 
 		/* Process all indexes in a single loop */
-		ReindexIndexesConcurrently(idxinfos, heaprels, params->options, reindex_context);
+		ReindexIndexesConcurrently(idxinfos, params->options, reindex_context);
 	} else {
 		/*
 		 * Process each partition listed in a separate transaction.  Note that
@@ -3262,7 +3257,6 @@ ReindexMultipleInternal(List *relids, ReindexParams *params)
 static bool
 ReindexRelationConcurrently(Oid relationOid, ReindexParams *params)
 {
-	List	   *heapRelationIds = NIL;
 	List	   *indexIds = NIL;
 	List	   *newIndexIds = NIL;
 	ListCell   *lc,
@@ -3315,14 +3309,6 @@ ReindexRelationConcurrently(Oid relationOid, ReindexParams *params)
 				 */
 				Relation	heapRelation;
 
-				/* Save the list of relation OIDs in private context */
-				oldcontext = MemoryContextSwitchTo(private_context);
-
-				/* Track this relation for session locks */
-				heapRelationIds = lappend_oid(heapRelationIds, relationOid);
-
-				MemoryContextSwitchTo(oldcontext);
-
 				if (IsCatalogRelationOid(relationOid))
 					ereport(ERROR,
 							(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
@@ -3335,7 +3321,7 @@ ReindexRelationConcurrently(Oid relationOid, ReindexParams *params)
 												  ShareUpdateExclusiveLock);
 					/* leave if relation does not exist */
 					if (!heapRelation)
-						break;
+						break; // XXX: lremove
 				}
 				else
 					heapRelation = table_open(relationOid,
@@ -3386,14 +3372,6 @@ ReindexRelationConcurrently(Oid relationOid, ReindexParams *params)
 					Relation	toastRelation = table_open(toastOid,
 														   ShareUpdateExclusiveLock);
 
-					/* Save the list of relation OIDs in private context */
-					oldcontext = MemoryContextSwitchTo(private_context);
-
-					/* Track this relation for session locks */
-					heapRelationIds = lappend_oid(heapRelationIds, toastOid);
-
-					MemoryContextSwitchTo(oldcontext);
-
 					foreach(lc2, RelationGetIndexList(toastRelation))
 					{
 						Oid			cellOid = lfirst_oid(lc2);
@@ -3434,70 +3412,6 @@ ReindexRelationConcurrently(Oid relationOid, ReindexParams *params)
 				break;
 			}
 		case RELKIND_INDEX:
-			{
-				Oid			heapId = IndexGetRelation(relationOid,
-													  (params->options & REINDEXOPT_MISSING_OK) != 0);
-				Relation	heapRelation;
-				ReindexIndexInfo *idx;
-
-				/* if relation is missing, leave */
-				if (!OidIsValid(heapId))
-					break;
-
-				if (IsCatalogRelationOid(heapId))
-					ereport(ERROR,
-							(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-							 errmsg("cannot reindex system catalogs concurrently")));
-
-				/*
-				 * Don't allow reindex for an invalid index on TOAST table, as
-				 * if rebuilt it would not be possible to drop it.  Match
-				 * error message in reindex_index().
-				 */
-				if (IsToastNamespace(get_rel_namespace(relationOid)) &&
-					!get_index_isvalid(relationOid))
-					ereport(ERROR,
-							(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-							 errmsg("cannot reindex invalid index on TOAST table")));
-
-				/*
-				 * Check if parent relation can be locked and if it exists,
-				 * this needs to be done at this stage as the list of indexes
-				 * to rebuild is not complete yet, and REINDEXOPT_MISSING_OK
-				 * should not be used once all the session locks are taken.
-				 */
-				if ((params->options & REINDEXOPT_MISSING_OK) != 0)
-				{
-					heapRelation = try_table_open(heapId,
-												  ShareUpdateExclusiveLock);
-					/* leave if relation does not exist */
-					if (!heapRelation)
-						break;
-				}
-				else
-					heapRelation = table_open(heapId,
-											  ShareUpdateExclusiveLock);
-				table_close(heapRelation, NoLock);
-
-				/* Save the list of relation OIDs in private context */
-				oldcontext = MemoryContextSwitchTo(private_context);
-
-				/* Track the heap relation of this index for session locks */
-				heapRelationIds = list_make1_oid(heapId);
-
-				/*
-				 * Save the list of relation OIDs in private context.  Note
-				 * that invalid indexes are allowed here.
-				 */
-				idx = palloc(sizeof(ReindexIndexInfo));
-				idx->indexId = relationOid;
-				indexIds = lappend(indexIds, idx);
-				/* other fields set later */
-
-				MemoryContextSwitchTo(oldcontext);
-				break;
-			}
-
 		case RELKIND_PARTITIONED_TABLE:
 		case RELKIND_PARTITIONED_INDEX:
 		default:
@@ -3521,10 +3435,8 @@ ReindexRelationConcurrently(Oid relationOid, ReindexParams *params)
 		return false;
 	}
 
-	Assert(heapRelationIds != NIL);
-
 	/* Do the work */
-	newIndexIds = ReindexIndexesConcurrently(indexIds, heapRelationIds, params->options, private_context);
+	newIndexIds = ReindexIndexesConcurrently(indexIds, params->options, private_context);
 
 	/* Log what we did */
 	if ((params->options & REINDEXOPT_VERBOSE) != 0)
@@ -3566,9 +3478,10 @@ ReindexRelationConcurrently(Oid relationOid, ReindexParams *params)
  * This is called by ReindexRelationConcurrently and
  */
 static List *
-ReindexIndexesConcurrently(List *indexIds, List *heapRelationIds, int options,
+ReindexIndexesConcurrently(List *indexIds, int options,
 		MemoryContext private_context)
 {
+	List		*heapRelationIds = NIL;
 	List	   *newIndexIds = NIL;
 	List	   *relationLocks = NIL;
 	List	   *lockTags = NIL;
@@ -3586,6 +3499,72 @@ ReindexIndexesConcurrently(List *indexIds, List *heapRelationIds, int options,
 	};
 	int64		progress_vals[4];
 
+	foreach(lc, indexIds)
+	{
+		ReindexIndexInfo	*idx = lfirst(lc);
+		Oid			indexrelid = idx->indexId;
+		Oid			heapId = IndexGetRelation(indexrelid,
+											  (options & REINDEXOPT_MISSING_OK) != 0);
+		Relation	heapRelation;
+
+		/* if relation is missing, leave */
+		if (!OidIsValid(heapId))
+			break; // XXX: ldelete?
+
+		if (IsCatalogRelationOid(heapId))
+			ereport(ERROR,
+					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+					 errmsg("cannot reindex system catalogs concurrently")));
+
+		/*
+		 * Don't allow reindex for an invalid index on TOAST table, as
+		 * if rebuilt it would not be possible to drop it.  Match
+		 * error message in reindex_index().
+		 */
+		if (IsToastNamespace(get_rel_namespace(indexrelid)) &&
+			!get_index_isvalid(indexrelid))
+			ereport(ERROR,
+					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+					 errmsg("cannot reindex invalid index on TOAST table")));
+
+		if (IsToastNamespace(get_rel_namespace(indexrelid)) &&
+			!get_index_isvalid(indexrelid))
+			ereport(ERROR,
+					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
+					 errmsg("cannot reindex invalid index on TOAST table")));
+
+		/*
+		 * Check if parent relation can be locked and if it exists,
+		 * this needs to be done at this stage as the list of indexes
+		 * to rebuild is not complete yet, and REINDEXOPT_MISSING_OK
+		 * should not be used once all the session locks are taken.
+		 */
+		if ((options & REINDEXOPT_MISSING_OK) != 0)
+		{
+			heapRelation = try_table_open(heapId,
+										  ShareUpdateExclusiveLock);
+			/* leave if relation does not exist */
+			if (!heapRelation)
+				break; // ldelete
+		}
+		else
+			heapRelation = table_open(heapId,
+									  ShareUpdateExclusiveLock);
+		table_close(heapRelation, NoLock);
+
+		/* Save the list of relation OIDs in private context */
+		oldcontext = MemoryContextSwitchTo(private_context);
+
+		/* Track the heap relation of this index for session locks */
+		heapRelationIds = lappend_oid(heapRelationIds, heapId);
+		// heapRelationIds = list_make1_oid(heapId);
+
+		/* Note that invalid indexes are allowed here. */
+
+		MemoryContextSwitchTo(oldcontext);
+		// break;
+	}
+
 	/*-----
 	 * Now we have all the indexes we want to process in indexIds.
 	 *
diff --git a/src/test/regress/expected/create_index.out b/src/test/regress/expected/create_index.out
index 4a03ab2abb..fc6afab58a 100644
--- a/src/test/regress/expected/create_index.out
+++ b/src/test/regress/expected/create_index.out
@@ -2470,12 +2470,12 @@ COMMIT;
 REINDEX TABLE CONCURRENTLY pg_class; -- no catalog relation
 ERROR:  cannot reindex system catalogs concurrently
 REINDEX INDEX CONCURRENTLY pg_class_oid_index; -- no catalog index
-ERROR:  concurrent index creation on system catalog tables is not supported
+ERROR:  cannot reindex system catalogs concurrently
 -- These are the toast table and index of pg_authid.
 REINDEX TABLE CONCURRENTLY pg_toast.pg_toast_1260; -- no catalog toast table
 ERROR:  cannot reindex system catalogs concurrently
 REINDEX INDEX CONCURRENTLY pg_toast.pg_toast_1260_index; -- no catalog toast index
-ERROR:  concurrent index creation on system catalog tables is not supported
+ERROR:  cannot reindex system catalogs concurrently
 REINDEX SYSTEM CONCURRENTLY postgres; -- not allowed for SYSTEM
 ERROR:  cannot reindex system catalogs concurrently
 -- Warns about catalog relations
-- 
2.17.0


--MfFXiAuoTsnnDAfZ--





^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44 Andres Freund <andres@anarazel.de>
  0 siblings, 0 replies; 325+ messages in thread

From: Andres Freund @ 2025-03-07 14:44 UTC (permalink / raw)

The test occasionally failed due to unexpected connection limit errors being
encountered after having waited for FATAL errors on another connection. These
spurious failures were caused by the the backend reporting FATAL errors to the
client before detaching from the PGPROC entry. Adding a sleep(1) before
proc_exit() makes it easy to reproduce that problem.

To fix the issue, add a helper function that waits for postmaster to notice
the process having exited. For now this is implemented by waiting for the
DEBUG2 message that postmaster logs in that case. That's not the prettiest
fix, but simple. If we notice this problem elsewhere, it might be worthwhile
to make this more general, e.g. by adding an injection point.

Reported-by: Tomas Vondra <tomas@vondra.me>
Diagnosed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/ggflhkciwdyotpoie323chu2c2idpjk5qimrn462encwx2io7s@thmcxl7i6dpw
---
 .../postmaster/t/002_connection_limits.pl     | 35 +++++++++++++++++--
 1 file changed, 32 insertions(+), 3 deletions(-)

diff --git a/src/test/postmaster/t/002_connection_limits.pl b/src/test/postmaster/t/002_connection_limits.pl
index 8cfa6e0ced5..2c185eef6eb 100644
--- a/src/test/postmaster/t/002_connection_limits.pl
+++ b/src/test/postmaster/t/002_connection_limits.pl
@@ -20,6 +20,7 @@ $node->append_conf('postgresql.conf', "max_connections = 6");
 $node->append_conf('postgresql.conf', "reserved_connections = 2");
 $node->append_conf('postgresql.conf', "superuser_reserved_connections = 1");
 $node->append_conf('postgresql.conf', "log_connections = on");
+$node->append_conf('postgresql.conf', "log_min_messages=debug2");
 $node->start;
 
 $node->safe_psql(
@@ -45,13 +46,39 @@ sub background_psql_as_user
 		extra_params => [ '-U', $user ]);
 }
 
+# Like connect_fails(), except that we also wait for the failed backend to
+# have exited.
+#
+# This tests needs to wait for client processes to exit because the error
+# message for a failed connection is reported before the backend has detached
+# from shared memory. If we didn't wait, subsequent tests might hit connection
+# limits spuriously.
+#
+# This can't easily be generalized, as detecting process exit requires
+# log_min_messages to be at least DEBUG2 and is not concurrency safe, as we
+# can't easily be sure the right process exited. In this test that's not a
+# problem though, we only have one new connection at a time.
+sub connect_fails_wait
+{
+	local $Test::Builder::Level = $Test::Builder::Level + 1;
+	my ($node, $connstr, $test_name, %params) = @_;
+
+	my $log_location = -s $node->logfile;
+
+	$node->connect_fails($connstr, $test_name, %params);
+	$node->wait_for_log(qr/DEBUG:  client backend.*exited with exit code 1/,
+		$log_location);
+	ok(1, "$test_name: client backend process exited");
+}
+
 my @sessions = ();
 my @raw_connections = ();
 
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
 push(@sessions, background_psql_as_user('regress_regular'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -60,7 +87,8 @@ $node->connect_fails(
 
 push(@sessions, background_psql_as_user('regress_reserved'));
 push(@sessions, background_psql_as_user('regress_reserved'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_regular",
 	"reserved_connections limit",
 	expected_stderr =>
@@ -68,7 +96,8 @@ $node->connect_fails(
 );
 
 push(@sessions, background_psql_as_user('regress_superuser'));
-$node->connect_fails(
+connect_fails_wait(
+	$node,
 	"dbname=postgres user=regress_superuser",
 	"superuser_reserved_connections limit",
 	expected_stderr => qr/FATAL:  sorry, too many clients already/);
-- 
2.48.1.76.g4e746b1a31.dirty


--q4d36jgrlhdhudrc
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment;
	filename="v2-0004-tests-Don-t-fail-due-to-high-default-timeout-in-p.patch"



^ permalink  raw  reply  [nested|flat] 325+ messages in thread


end of thread, other threads:[~2025-03-07 14:44 UTC | newest]

Thread overview: 325+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2020-11-01 19:46 [PATCH v12 5/5] More refactoring Justin Pryzby <pryzbyj@telsasoft.com>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>
2025-03-07 14:44 [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits Andres Freund <andres@anarazel.de>

This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox