Stabilize 019_replslot_limit.pl: wait on slot restart_lsn

wait_for_catchup() has "wait for the standby to reach the target LSN"
semantics.  However, the previous polling implementation actually waited for
the primary to observe that position via pg_stat_replication.
7e8aeb9e483 introduced the new WAIT FOR LSN-based implementation, which
just probes the standby.

019_replslot_limit.pl relied on the old side effect: its
"slot state changes to extended/unreserved" subtests inspect
primary-side pg_replication_slots, whose wal_status depends on
restart_lsn, which only advances after the walsender processes a
standby reply.  Make the test wait on what it actually needs by
replacing each wait_for_catchup() with
wait_for_slot_catchup('rep1', 'restart', primary->lsn('write')).

Reported-by: Alexander Lakhin <exclusion@gmail.com>
Discussion: https://postgr.es/m/63f6abc9-c0ae-465d-a4e6-667eca6ea008@gmail.com
Author: Xuneng Zhou <xunengzhou@gmail.com>
Author: Alexander Korotkov <aekorotkov@gmail.com>

diff --git a/src/test/recovery/t/019_replslot_limit.pl b/src/test/recovery/t/019_replslot_limit.pl
index 7b253e64d9ccbcc7edc88abcd9412af61ccc8bbc..882ffb66550148bf4b19b4625742fe4356cffdf6 100644 (file)
--- a/src/test/recovery/t/019_replslot_limit.pl
+++ b/src/test/recovery/t/019_replslot_limit.pl
@@ -44,8 +44,12 @@ $node_standby->append_conf('postgresql.conf', "primary_slot_name = 'rep1'");
 
 $node_standby->start;
 
-# Wait until standby has replayed enough data
-$node_primary->wait_for_catchup($node_standby);
+# Wait until the primary has processed standby feedback and advanced the
+# slot's restart_lsn.  For a physical slot, restart_lsn is updated from
+# the standby's reported flush position, so this waits for the primary-side
+# slot state that the following wal_status checks depend on.
+$node_primary->wait_for_slot_catchup('rep1', 'restart',
+       $node_primary->lsn('write'));
 
 # Stop standby
 $node_standby->stop;
@@ -79,7 +83,8 @@ is($result, "reserved|t", 'check that slot is working');
 # The standby can reconnect to primary
 $node_standby->start;
 
-$node_primary->wait_for_catchup($node_standby);
+$node_primary->wait_for_slot_catchup('rep1', 'restart',
+       $node_primary->lsn('write'));
 
 $node_standby->stop;
 
@@ -109,7 +114,8 @@ is($result, "reserved",
 
 # The standby can reconnect to primary
 $node_standby->start;
-$node_primary->wait_for_catchup($node_standby);
+$node_primary->wait_for_slot_catchup('rep1', 'restart',
+       $node_primary->lsn('write'));
 $node_standby->stop;
 
 # wal_keep_size overrides max_slot_wal_keep_size
@@ -128,7 +134,8 @@ $result = $node_primary->safe_psql('postgres',
 
 # The standby can reconnect to primary
 $node_standby->start;
-$node_primary->wait_for_catchup($node_standby);
+$node_primary->wait_for_slot_catchup('rep1', 'restart',
+       $node_primary->lsn('write'));
 $node_standby->stop;
 
 # Advance WAL again without checkpoint, reducing remain by 6 MB.
@@ -155,7 +162,8 @@ is($result, "unreserved|t",
 # The standby still can connect to primary before a checkpoint
 $node_standby->start;
 
-$node_primary->wait_for_catchup($node_standby);
+$node_primary->wait_for_slot_catchup('rep1', 'restart',
+       $node_primary->lsn('write'));
 
 $node_standby->stop;