Fix test race between primary XLOG_RUNNING_XACTS and standby logical slot.

Before the previous commit, the test could hang until
LOG_SNAPSHOT_INTERVAL_MS (15s), until checkpoint_timeout (300s), or
indefinitely.  An indefinite hang was awfully improbable.  It entailed
the test reaching checkpoint_timeout before the
DecodingContextFindStartpoint() of a CREATE SUBSCRIPTION, yet after the
preceding WAL record.  Back-patch to v16, which introduced the test.

Bertrand Drouvot, reported by Noah Misch.

Discussion: https://postgr.es/m/20240211010227.a2.nmisch@google.com

diff --git a/src/test/perl/PostgreSQL/Test/Cluster.pm b/src/test/perl/PostgreSQL/Test/Cluster.pm
index 5f0f04b55583cc2126d5e58edcc48b7b40b7fa61..7743ffd544d57a77f6ce9e495d0e6080796dfcb1 100644 (file)
--- a/src/test/perl/PostgreSQL/Test/Cluster.pm
+++ b/src/test/perl/PostgreSQL/Test/Cluster.pm
@@ -3092,6 +3092,36 @@ $SIG{TERM} = $SIG{INT} = sub {
 
 =pod
 
+=item $node->log_standby_snapshot(self, standby, slot_name)
+
+Log a standby snapshot on primary once the slot restart_lsn is determined on
+the standby.
+
+=cut
+
+sub log_standby_snapshot
+{
+   my ($self, $standby, $slot_name) = @_;
+
+   # Once the slot's restart_lsn is determined, the standby looks for
+   # xl_running_xacts WAL record from the restart_lsn onwards. First wait
+   # until the slot restart_lsn is determined.
+
+   $standby->poll_query_until(
+       'postgres', qq[
+       SELECT restart_lsn IS NOT NULL
+       FROM pg_catalog.pg_replication_slots WHERE slot_name = '$slot_name'
+   ])
+     or die
+     "timed out waiting for logical slot to calculate its restart_lsn";
+
+   # Then arrange for the xl_running_xacts record for which the standby is
+   # waiting.
+   $self->safe_psql('postgres', 'SELECT pg_log_standby_snapshot()');
+}
+
+=pod
+
 =item $node->create_logical_slot_on_standby(self, primary, slot_name, dbname)
 
 Create logical replication slot on given standby
@@ -3117,21 +3147,9 @@ sub create_logical_slot_on_standby
        '2>',
        \$stderr);
 
-   # Once the slot's restart_lsn is determined, the standby looks for
-   # xl_running_xacts WAL record from the restart_lsn onwards. First wait
-   # until the slot restart_lsn is determined.
-
-   $self->poll_query_until(
-       'postgres', qq[
-       SELECT restart_lsn IS NOT NULL
-       FROM pg_catalog.pg_replication_slots WHERE slot_name = '$slot_name'
-   ])
-     or die
-     "timed out waiting for logical slot to calculate its restart_lsn";
-
-   # Then arrange for the xl_running_xacts record for which pg_recvlogical is
+   # Arrange for the xl_running_xacts record for which pg_recvlogical is
    # waiting.
-   $primary->safe_psql('postgres', 'SELECT pg_log_standby_snapshot()');
+   $primary->log_standby_snapshot($self, $slot_name);
 
    $handle->finish();
 

diff --git a/src/test/recovery/t/035_standby_logical_decoding.pl b/src/test/recovery/t/035_standby_logical_decoding.pl
index f02eb39de929d6a929c0a31e61463a248a169b67..e90191fe2aaa1fde961f4a7ddcc885e49f4dd2c5 100644 (file)
--- a/src/test/recovery/t/035_standby_logical_decoding.pl
+++ b/src/test/recovery/t/035_standby_logical_decoding.pl
@@ -467,8 +467,8 @@ $psql_subscriber{subscriber_stdin} .= "\n";
 
 $psql_subscriber{run}->pump_nb();
 
-# Speed up the subscription creation
-$node_primary->safe_psql('postgres', "SELECT pg_log_standby_snapshot()");
+# Log the standby snapshot to speed up the subscription creation
+$node_primary->log_standby_snapshot($node_standby, 'tap_sub');
 
 # Explicitly shut down psql instance gracefully - to avoid hangs
 # or worse on windows