| 1 |
|
| 2 |
# Copyright (c) 2021-2025, PostgreSQL Global Development Group |
| 3 |
|
| 4 |
# Test for checking consistency of on-disk pages for a cluster with |
| 5 |
# the minimum recovery LSN, ensuring that the updates happen across |
| 6 |
# all processes. In this test, the updates from the startup process |
| 7 |
# and the checkpointer (which triggers non-startup code paths) are |
| 8 |
# both checked. |
| 9 |
|
| 10 |
use strict; |
| 11 |
use warnings FATAL => 'all'; |
| 12 |
use PostgreSQL::Test::Cluster; |
| 13 |
use PostgreSQL::Test::Utils; |
| 14 |
use Test::More; |
| 15 |
|
| 16 |
# Find the largest LSN in the set of pages part of the given relation |
| 17 |
# file. This is used for offline checks of page consistency. The LSN |
| 18 |
# is historically stored as a set of two numbers of 4 byte-length |
| 19 |
# located at the beginning of each page. |
| 20 |
sub find_largest_lsn |
| 21 |
{ |
| 22 |
my $blocksize = int(shift); |
| 23 |
my $filename = shift; |
| 24 |
my ($max_hi, $max_lo) = (0, 0); |
| 25 |
open(my $fh, "<:raw", $filename) |
| 26 |
or die "failed to open $filename: $!"; |
| 27 |
my ($buf, $len); |
| 28 |
while ($len = read($fh, $buf, $blocksize)) |
| 29 |
{ |
| 30 |
$len == $blocksize |
| 31 |
or die "read only $len of $blocksize bytes from $filename"; |
| 32 |
my ($hi, $lo) = unpack("LL", $buf); |
| 33 |
|
| 34 |
if ($hi > $max_hi or ($hi == $max_hi and $lo > $max_lo)) |
| 35 |
{ |
| 36 |
($max_hi, $max_lo) = ($hi, $lo); |
| 37 |
} |
| 38 |
} |
| 39 |
defined($len) or die "read error on $filename: $!"; |
| 40 |
close($fh); |
| 41 |
|
| 42 |
return sprintf("%X/%X", $max_hi, $max_lo); |
| 43 |
} |
| 44 |
|
| 45 |
# Initialize primary node |
| 46 |
my $primary = PostgreSQL::Test::Cluster->new('primary'); |
| 47 |
$primary->init(allows_streaming => 1); |
| 48 |
|
| 49 |
# Set shared_buffers to a very low value to enforce discard and flush |
| 50 |
# of PostgreSQL buffers on standby, enforcing other processes than the |
| 51 |
# startup process to update the minimum recovery LSN in the control |
| 52 |
# file. Autovacuum is disabled so as there is no risk of having other |
| 53 |
# processes than the checkpointer doing page flushes. |
| 54 |
$primary->append_conf("postgresql.conf", <<EOF); |
| 55 |
shared_buffers = 128kB |
| 56 |
autovacuum = off |
| 57 |
EOF |
| 58 |
|
| 59 |
# Start the primary |
| 60 |
$primary->start; |
| 61 |
|
| 62 |
# setup/start a standby |
| 63 |
$primary->backup('bkp'); |
| 64 |
my $standby = PostgreSQL::Test::Cluster->new('standby'); |
| 65 |
$standby->init_from_backup($primary, 'bkp', has_streaming => 1); |
| 66 |
$standby->start; |
| 67 |
|
| 68 |
# Create base table whose data consistency is checked. |
| 69 |
$primary->safe_psql( |
| 70 |
'postgres', " |
| 71 |
CREATE TABLE test1 (a int) WITH (fillfactor = 10); |
| 72 |
INSERT INTO test1 SELECT generate_series(1, 10000);"); |
| 73 |
|
| 74 |
# Take a checkpoint and enforce post-checkpoint full page writes |
| 75 |
# which makes the startup process replay those pages, updating |
| 76 |
# minRecoveryPoint. |
| 77 |
$primary->safe_psql('postgres', 'CHECKPOINT;'); |
| 78 |
$primary->safe_psql('postgres', 'UPDATE test1 SET a = a + 1;'); |
| 79 |
|
| 80 |
# Wait for last record to have been replayed on the standby. |
| 81 |
$primary->wait_for_catchup($standby); |
| 82 |
|
| 83 |
# Fill in the standby's shared buffers with the data filled in |
| 84 |
# previously. |
| 85 |
$standby->safe_psql('postgres', 'SELECT count(*) FROM test1;'); |
| 86 |
|
| 87 |
# Update the table again, this does not generate full page writes so |
| 88 |
# the standby will replay records associated with it, but the startup |
| 89 |
# process will not flush those pages. |
| 90 |
$primary->safe_psql('postgres', 'UPDATE test1 SET a = a + 1;'); |
| 91 |
|
| 92 |
# Extract from the relation the last block created and its relation |
| 93 |
# file, this will be used at the end of the test for sanity checks. |
| 94 |
my $blocksize = $primary->safe_psql('postgres', |
| 95 |
"SELECT setting::int FROM pg_settings WHERE name = 'block_size';"); |
| 96 |
my $last_block = $primary->safe_psql('postgres', |
| 97 |
"SELECT pg_relation_size('test1')::int / $blocksize - 1;"); |
| 98 |
my $relfilenode = $primary->safe_psql('postgres', |
| 99 |
"SELECT pg_relation_filepath('test1'::regclass);"); |
| 100 |
|
| 101 |
# Wait for last record to have been replayed on the standby. |
| 102 |
$primary->wait_for_catchup($standby); |
| 103 |
|
| 104 |
# Issue a restart point on the standby now, which makes the checkpointer |
| 105 |
# update minRecoveryPoint. |
| 106 |
$standby->safe_psql('postgres', 'CHECKPOINT;'); |
| 107 |
|
| 108 |
# Now shut down the primary violently so as the standby does not |
| 109 |
# receive the shutdown checkpoint, making sure that the startup |
| 110 |
# process does not flush any pages on its side. The standby is |
| 111 |
# cleanly stopped, which makes the checkpointer update minRecoveryPoint |
| 112 |
# with the restart point created at shutdown. |
| 113 |
$primary->stop('immediate'); |
| 114 |
$standby->stop('fast'); |
| 115 |
|
| 116 |
# Check the data consistency of the instance while offline. This is |
| 117 |
# done by directly scanning the on-disk relation blocks and what |
| 118 |
# pg_controldata lets know. |
| 119 |
my $standby_data = $standby->data_dir; |
| 120 |
my $offline_max_lsn = |
| 121 |
find_largest_lsn($blocksize, "$standby_data/$relfilenode"); |
| 122 |
|
| 123 |
# Fetch minRecoveryPoint from the control file itself |
| 124 |
my ($stdout, $stderr) = run_command([ 'pg_controldata', $standby_data ]); |
| 125 |
my @control_data = split("\n", $stdout); |
| 126 |
my $offline_recovery_lsn = undef; |
| 127 |
foreach (@control_data) |
| 128 |
{ |
| 129 |
if ($_ =~ /^Minimum recovery ending location:\s*(.*)$/mg) |
| 130 |
{ |
| 131 |
$offline_recovery_lsn = $1; |
| 132 |
last; |
| 133 |
} |
| 134 |
} |
| 135 |
die "No minRecoveryPoint in control file found\n" |
| 136 |
unless defined($offline_recovery_lsn); |
| 137 |
|
| 138 |
# minRecoveryPoint should never be older than the maximum LSN for all |
| 139 |
# the pages on disk. |
| 140 |
ok($offline_recovery_lsn ge $offline_max_lsn, |
| 141 |
"Check offline that table data is consistent with minRecoveryPoint"); |
| 142 |
|
| 143 |
done_testing(); |