]> git.ipfire.org Git - thirdparty/postgresql.git/commit
Fix assertion failure in snapshot building
authorDaniel Gustafsson <dgustafsson@postgresql.org>
Tue, 4 Jul 2023 15:36:13 +0000 (17:36 +0200)
committerDaniel Gustafsson <dgustafsson@postgresql.org>
Tue, 4 Jul 2023 15:36:13 +0000 (17:36 +0200)
commit48efb2302bea3b2af8b4bb865edcc70132ee4250
treea8a46c0ec38a78ce7fe0d4caf2a77116e373e1de
parent4b4798e1384ce12e19a558708e66c829276cf8d7
Fix assertion failure in snapshot building

Clear any potential stale next_phase_at value from the snapshot
builder which otherwise may trip an assertion check ensuring
that there is no next_phase_at value.

This can be reproduced by running 80 concurrent sessions like
the below where $c is a loop counter (assumes there has been
1..$c databases created) :

  echo "
    CREATE TABLE replication_example(id SERIAL PRIMARY KEY,
                                     somedata int,
                                     text varchar(120));
    SELECT 'init' FROM
      pg_create_logical_replication_slot('regression_slot_$c',
                                         'test_decoding');
    SELECT data FROM
      pg_logical_slot_get_changes('regression_slot_$c', NULL,
                                  NULL, 'include-xids', '0',
                                  'skip-empty-xacts', '1');
  " | psql -d regress_$c >>psql.log &

Backpatch down to v16.

Bug: #17695
Author: Masahiko Sawada <sawada.mshk@gmail.com>
Reviewed-by: Alexander Lakhin <exclusion@gmail.com>
Reported-by: bowenshi <zxwsbg@qq.com>
Discussion: https://postgr.es/m/17695-6be9277c9295985f@postgresql.org
Backpatch-through: v16
src/backend/replication/logical/snapbuild.c