agora inbox for [email protected]  

help / color / mirror / Atom feed

[PATCH v29 03/11] Allow to prolong life span of transition tables until transaction end
126+ messages / 2 participants
[nested] [flat]

* [PATCH v29 03/11] Allow to prolong life span of transition tables until transaction end
@ 2019-12-20 01:09  Yugo Nagata <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

From: Yugo Nagata @ 2019-12-20 01:09 UTC (permalink / raw)

Originally, tuplestores of AFTER trigger's transition tables were
freed for each query depth. For our IVM implementation, we would like
to prolong life of the tuplestores because we have to preserve them
for a whole query assuming that some base tables might be changed
in some trigger functions.
---
 src/backend/commands/trigger.c | 83 ++++++++++++++++++++++++++++++++--
 src/include/commands/trigger.h |  2 +
 2 files changed, 81 insertions(+), 4 deletions(-)

diff --git a/src/backend/commands/trigger.c b/src/backend/commands/trigger.c
index 52177759ab..00b20f4b5b 100644
--- a/src/backend/commands/trigger.c
+++ b/src/backend/commands/trigger.c
@@ -3742,6 +3742,10 @@ typedef struct AfterTriggerEventList
  * end of the list, so it is relatively easy to discard them.  The event
  * list chunks themselves are stored in event_cxt.
  *
+ * prolonged_tuplestored is a list of transition table tuplestores whose
+ * life are prolonged to the end of the outmost query instead of each nested
+ * query.
+ *
  * query_depth is the current depth of nested AfterTriggerBeginQuery calls
  * (-1 when the stack is empty).
  *
@@ -3807,6 +3811,7 @@ typedef struct AfterTriggersData
 	SetConstraintState state;	/* the active S C state */
 	AfterTriggerEventList events;	/* deferred-event list */
 	MemoryContext event_cxt;	/* memory context for events, if any */
+	List   *prolonged_tuplestores;	/* list of prolonged tuplestores */
 
 	/* per-query-level data: */
 	AfterTriggersQueryData *query_stack;	/* array of structs shown below */
@@ -3842,6 +3847,7 @@ struct AfterTriggersTableData
 	bool		closed;			/* true when no longer OK to add tuples */
 	bool		before_trig_done;	/* did we already queue BS triggers? */
 	bool		after_trig_done;	/* did we already queue AS triggers? */
+	bool		prolonged;			/* are transition tables prolonged? */
 	AfterTriggerEventList after_trig_events;	/* if so, saved list pointer */
 
 	/*
@@ -3891,6 +3897,7 @@ static void TransitionTableAddTuple(EState *estate,
 									TupleTableSlot *original_insert_tuple,
 									Tuplestorestate *tuplestore);
 static void AfterTriggerFreeQuery(AfterTriggersQueryData *qs);
+static void release_or_prolong_tuplestore(Tuplestorestate *ts, bool prolonged);
 static SetConstraintState SetConstraintStateCreate(int numalloc);
 static SetConstraintState SetConstraintStateCopy(SetConstraintState origstate);
 static SetConstraintState SetConstraintStateAddItem(SetConstraintState state,
@@ -4768,6 +4775,45 @@ afterTriggerInvokeEvents(AfterTriggerEventList *events,
 }
 
 
+/*
+ * SetTransitionTablePreserved
+ *
+ * Prolong lifespan of transition tables corresponding specified relid and
+ * command type to the end of the outmost query instead of each nested query.
+ * This enables to use nested AFTER trigger's transition tables from outer
+ * query's triggers.  Currently, only immediate incremental view maintenance
+ * uses this.
+ */
+void
+SetTransitionTablePreserved(Oid relid, CmdType cmdType)
+{
+	AfterTriggersTableData *table;
+	AfterTriggersQueryData *qs;
+	bool		found = false;
+	ListCell   *lc;
+
+	/* Check state, like AfterTriggerSaveEvent. */
+	if (afterTriggers.query_depth < 0)
+		elog(ERROR, "SetTransitionTablePreserved() called outside of query");
+
+	qs = &afterTriggers.query_stack[afterTriggers.query_depth];
+
+	foreach(lc, qs->tables)
+	{
+		table = (AfterTriggersTableData *) lfirst(lc);
+		if (table->relid == relid && table->cmdType == cmdType &&
+			table->closed)
+		{
+			table->prolonged = true;
+			found = true;
+		}
+	}
+
+	if (!found)
+		elog(ERROR,"could not find table with OID %d and command type %d", relid, cmdType);
+}
+
+
 /*
  * GetAfterTriggersTableData
  *
@@ -4978,6 +5024,7 @@ AfterTriggerBeginXact(void)
 	 */
 	afterTriggers.firing_counter = (CommandId) 1;	/* mustn't be 0 */
 	afterTriggers.query_depth = -1;
+	afterTriggers.prolonged_tuplestores = NIL;
 
 	/*
 	 * Verify that there is no leftover state remaining.  If these assertions
@@ -5138,19 +5185,19 @@ AfterTriggerFreeQuery(AfterTriggersQueryData *qs)
 		ts = table->old_upd_tuplestore;
 		table->old_upd_tuplestore = NULL;
 		if (ts)
-			tuplestore_end(ts);
+			release_or_prolong_tuplestore(ts, table->prolonged);
 		ts = table->new_upd_tuplestore;
 		table->new_upd_tuplestore = NULL;
 		if (ts)
-			tuplestore_end(ts);
+			release_or_prolong_tuplestore(ts, table->prolonged);
 		ts = table->old_del_tuplestore;
 		table->old_del_tuplestore = NULL;
 		if (ts)
-			tuplestore_end(ts);
+			release_or_prolong_tuplestore(ts, table->prolonged);
 		ts = table->new_ins_tuplestore;
 		table->new_ins_tuplestore = NULL;
 		if (ts)
-			tuplestore_end(ts);
+			release_or_prolong_tuplestore(ts, table->prolonged);
 		if (table->storeslot)
 		{
 			TupleTableSlot *slot = table->storeslot;
@@ -5167,6 +5214,34 @@ AfterTriggerFreeQuery(AfterTriggersQueryData *qs)
 	 */
 	qs->tables = NIL;
 	list_free_deep(tables);
+
+	/* Release prolonged tuplestores at the end of the outmost query */
+	if (afterTriggers.query_depth == 0)
+	{
+		foreach(lc, afterTriggers.prolonged_tuplestores)
+		{
+			ts = (Tuplestorestate *) lfirst(lc);
+			if (ts)
+				tuplestore_end(ts);
+		}
+		afterTriggers.prolonged_tuplestores = NIL;
+	}
+}
+
+/*
+ * Release the tuplestore, or append it to the prolonged tuplestores list.
+ */
+static void
+release_or_prolong_tuplestore(Tuplestorestate *ts, bool prolonged)
+{
+	if (prolonged && afterTriggers.query_depth > 0)
+	{
+		MemoryContext oldcxt = MemoryContextSwitchTo(CurTransactionContext);
+		afterTriggers.prolonged_tuplestores = lappend(afterTriggers.prolonged_tuplestores, ts);
+		MemoryContextSwitchTo(oldcxt);
+	}
+	else
+		tuplestore_end(ts);
 }
 
 
diff --git a/src/include/commands/trigger.h b/src/include/commands/trigger.h
index 430e3ca7dd..48a21c4c51 100644
--- a/src/include/commands/trigger.h
+++ b/src/include/commands/trigger.h
@@ -265,6 +265,8 @@ extern void AfterTriggerEndSubXact(bool isCommit);
 extern void AfterTriggerSetState(ConstraintsSetStmt *stmt);
 extern bool AfterTriggerPendingOnRel(Oid relid);
 
+extern void SetTransitionTablePreserved(Oid relid, CmdType cmdType);
+
 
 /*
  * in utils/adt/ri_triggers.c
-- 
2.25.1


--Multipart=_Mon__28_Aug_2023_11_52_52_+0900_hj6L5h176QaSGtg7
Content-Type: text/x-diff;
 name="v29-0004-Add-Incremental-View-Maintenance-support-to-pg_d.patch"
Content-Disposition: attachment;
 filename="v29-0004-Add-Incremental-View-Maintenance-support-to-pg_d.patch"
Content-Transfer-Encoding: 7bit



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread

* [PATCH v2 3/4] tests: Fix race condition in postmaster/002_connection_limits
@ 2025-03-07 14:44  Andres Freund <[email protected]>
  0 siblings, 0 replies; 126+ messages in thread

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

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

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

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

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


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



^ permalink  raw  reply  [nested|flat] 126+ messages in thread