| 1 |
|
| 2 |
# Copyright (c) 2021-2025, PostgreSQL Global Development Group |
| 3 |
|
| 4 |
# Checks for recovery_min_apply_delay and recovery pause |
| 5 |
use strict; |
| 6 |
use warnings FATAL => 'all'; |
| 7 |
|
| 8 |
use PostgreSQL::Test::Cluster; |
| 9 |
use PostgreSQL::Test::Utils; |
| 10 |
use Test::More; |
| 11 |
|
| 12 |
# Initialize primary node |
| 13 |
my $node_primary = PostgreSQL::Test::Cluster->new('primary'); |
| 14 |
$node_primary->init(allows_streaming => 1); |
| 15 |
$node_primary->start; |
| 16 |
|
| 17 |
# And some content |
| 18 |
$node_primary->safe_psql('postgres', |
| 19 |
"CREATE TABLE tab_int AS SELECT generate_series(1, 10) AS a"); |
| 20 |
|
| 21 |
# Take backup |
| 22 |
my $backup_name = 'my_backup'; |
| 23 |
$node_primary->backup($backup_name); |
| 24 |
|
| 25 |
# Create streaming standby from backup |
| 26 |
my $node_standby = PostgreSQL::Test::Cluster->new('standby'); |
| 27 |
my $delay = 3; |
| 28 |
$node_standby->init_from_backup($node_primary, $backup_name, |
| 29 |
has_streaming => 1); |
| 30 |
$node_standby->append_conf( |
| 31 |
'postgresql.conf', qq( |
| 32 |
recovery_min_apply_delay = '${delay}s' |
| 33 |
)); |
| 34 |
$node_standby->start; |
| 35 |
|
| 36 |
# Make new content on primary and check its presence in standby depending |
| 37 |
# on the delay applied above. Before doing the insertion, get the |
| 38 |
# current timestamp that will be used as a comparison base. Even on slow |
| 39 |
# machines, this allows to have a predictable behavior when comparing the |
| 40 |
# delay between data insertion moment on primary and replay time on standby. |
| 41 |
my $primary_insert_time = time(); |
| 42 |
$node_primary->safe_psql('postgres', |
| 43 |
"INSERT INTO tab_int VALUES (generate_series(11, 20))"); |
| 44 |
|
| 45 |
# Now wait for replay to complete on standby. We're done waiting when the |
| 46 |
# standby has replayed up to the previously saved primary LSN. |
| 47 |
my $until_lsn = |
| 48 |
$node_primary->safe_psql('postgres', "SELECT pg_current_wal_lsn()"); |
| 49 |
|
| 50 |
$node_standby->poll_query_until('postgres', |
| 51 |
"SELECT (pg_last_wal_replay_lsn() - '$until_lsn'::pg_lsn) >= 0") |
| 52 |
or die "standby never caught up"; |
| 53 |
|
| 54 |
# This test is successful if and only if the LSN has been applied with at least |
| 55 |
# the configured apply delay. |
| 56 |
ok(time() - $primary_insert_time >= $delay, |
| 57 |
"standby applies WAL only after replication delay"); |
| 58 |
|
| 59 |
|
| 60 |
# Check that recovery can be paused or resumed expectedly. |
| 61 |
my $node_standby2 = PostgreSQL::Test::Cluster->new('standby2'); |
| 62 |
$node_standby2->init_from_backup($node_primary, $backup_name, |
| 63 |
has_streaming => 1); |
| 64 |
$node_standby2->start; |
| 65 |
|
| 66 |
# Recovery is not yet paused. |
| 67 |
is( $node_standby2->safe_psql( |
| 68 |
'postgres', "SELECT pg_get_wal_replay_pause_state()"), |
| 69 |
'not paused', |
| 70 |
'pg_get_wal_replay_pause_state() reports not paused'); |
| 71 |
|
| 72 |
# Request to pause recovery and wait until it's actually paused. |
| 73 |
$node_standby2->safe_psql('postgres', "SELECT pg_wal_replay_pause()"); |
| 74 |
$node_primary->safe_psql('postgres', |
| 75 |
"INSERT INTO tab_int VALUES (generate_series(21,30))"); |
| 76 |
$node_standby2->poll_query_until('postgres', |
| 77 |
"SELECT pg_get_wal_replay_pause_state() = 'paused'") |
| 78 |
or die "Timed out while waiting for recovery to be paused"; |
| 79 |
|
| 80 |
# Even if new WAL records are streamed from the primary, |
| 81 |
# recovery in the paused state doesn't replay them. |
| 82 |
my $receive_lsn = |
| 83 |
$node_standby2->safe_psql('postgres', "SELECT pg_last_wal_receive_lsn()"); |
| 84 |
my $replay_lsn = |
| 85 |
$node_standby2->safe_psql('postgres', "SELECT pg_last_wal_replay_lsn()"); |
| 86 |
$node_primary->safe_psql('postgres', |
| 87 |
"INSERT INTO tab_int VALUES (generate_series(31,40))"); |
| 88 |
$node_standby2->poll_query_until('postgres', |
| 89 |
"SELECT '$receive_lsn'::pg_lsn < pg_last_wal_receive_lsn()") |
| 90 |
or die "Timed out while waiting for new WAL to be streamed"; |
| 91 |
is( $node_standby2->safe_psql('postgres', "SELECT pg_last_wal_replay_lsn()"), |
| 92 |
qq($replay_lsn), |
| 93 |
'no WAL is replayed in the paused state'); |
| 94 |
|
| 95 |
# Request to resume recovery and wait until it's actually resumed. |
| 96 |
$node_standby2->safe_psql('postgres', "SELECT pg_wal_replay_resume()"); |
| 97 |
$node_standby2->poll_query_until('postgres', |
| 98 |
"SELECT pg_get_wal_replay_pause_state() = 'not paused' AND pg_last_wal_replay_lsn() > '$replay_lsn'::pg_lsn" |
| 99 |
) or die "Timed out while waiting for recovery to be resumed"; |
| 100 |
|
| 101 |
# Check that the paused state ends and promotion continues if a promotion |
| 102 |
# is triggered while recovery is paused. |
| 103 |
$node_standby2->safe_psql('postgres', "SELECT pg_wal_replay_pause()"); |
| 104 |
$node_primary->safe_psql('postgres', |
| 105 |
"INSERT INTO tab_int VALUES (generate_series(41,50))"); |
| 106 |
$node_standby2->poll_query_until('postgres', |
| 107 |
"SELECT pg_get_wal_replay_pause_state() = 'paused'") |
| 108 |
or die "Timed out while waiting for recovery to be paused"; |
| 109 |
|
| 110 |
$node_standby2->promote; |
| 111 |
$node_standby2->poll_query_until('postgres', "SELECT NOT pg_is_in_recovery()") |
| 112 |
or die "Timed out while waiting for promotion to finish"; |
| 113 |
|
| 114 |
done_testing(); |