database/postgres
1
0
Fork 0
mirror of https://github.com/postgres/postgres.git synced 2025-07-28 23:42:10 +03:00
Code Activity

Restore replication protocol's duplicate command tags

Browse Source 
I removed the duplicate command tags for START_REPLICATION inadvertently
in commit 07082b08cc, but the replication protocol requires them.  The
fact that the replication protocol was broken was not noticed because
all our test cases use an optimized code path that exits early, failing
to verify that the behavior is correct for non-optimized cases.  Put
them back.

Also document this protocol quirk.

Add a test case that shows the failure.  It might still succeed even
without the patch when run on a fast enough server, but it suffices to
show the bug in enough cases that it would be noticed in buildfarm.

Author: Álvaro Herrera <alvherre@alvh.no-ip.org>
Reported-by: Henry Hinze <henry.hinze@gmail.com>
Reviewed-by: Petr Jelínek <petr.jelinek@2ndquadrant.com>
Discussion: https://postgr.es/m/16643-eaadeb2a1a58d28c@postgresql.org
This commit is contained in:
Alvaro Herrera
2020-10-14 20:12:26 -03:00
parent b94109ce37
commit 4e9821b6fa
4 changed files with 61 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
src/backend/replication/logical/worker.c
View File

@ -3071,7 +3071,6 @@ ApplyWorkerMain(Datum main_arg)
* does some initializations on the upstream so let's still call it.
*/
(void) walrcv_identify_system(wrconn, &startpointTLI);
}
/*

3
src/backend/replication/walsender.c
View File

@ -1656,7 +1656,8 @@ exec_replication_command(const char *cmd_string)
else
StartLogicalReplication(cmd);
/* callees already sent their own completion message */
/* dupe, but necessary per libpqrcv_endstreaming */
EndReplicationCommand(cmdtag);
Assert(xlogreader != NULL);
break;

55
src/test/subscription/t/100_bugs.pl
View File

@ -3,7 +3,7 @@ use strict;
use warnings;
use PostgresNode;
use TestLib;
use Test::More tests => 3;
use Test::More tests => 5;
# Bug #15114
@ -100,3 +100,56 @@ is( $node_publisher->psql(
);
$node_publisher->stop('fast');
# Bug #16643 - https://postgr.es/m/16643-eaadeb2a1a58d28c@postgresql.org
#
# Initial sync doesn't complete; the protocol was not being followed per
# expectations after commit 07082b08cc5d.
my $node_twoways = get_new_node('twoways');
$node_twoways->init(allows_streaming => 'logical');
$node_twoways->start;
for my $db (qw(d1 d2))
{
$node_twoways->safe_psql('postgres', "CREATE DATABASE $db");
$node_twoways->safe_psql($db, "CREATE TABLE t (f int)");
$node_twoways->safe_psql($db, "CREATE TABLE t2 (f int)");
}
my $rows = 3000;
$node_twoways->safe_psql(
'd1', qq{
INSERT INTO t SELECT * FROM generate_series(1, $rows);
INSERT INTO t2 SELECT * FROM generate_series(1, $rows);
CREATE PUBLICATION testpub FOR TABLE t;
SELECT pg_create_logical_replication_slot('testslot', 'pgoutput');
});
$node_twoways->safe_psql('d2',
"CREATE SUBSCRIPTION testsub CONNECTION \$\$"
. $node_twoways->connstr('d1')
. "\$\$ PUBLICATION testpub WITH (create_slot=false, "
. "slot_name='testslot')");
$node_twoways->safe_psql(
'd1', qq{
INSERT INTO t SELECT * FROM generate_series(1, $rows);
INSERT INTO t2 SELECT * FROM generate_series(1, $rows);
});
$node_twoways->safe_psql(
'd1', 'ALTER PUBLICATION testpub ADD TABLE t2');
$node_twoways->safe_psql(
'd2', 'ALTER SUBSCRIPTION testsub REFRESH PUBLICATION');
# We cannot rely solely on wait_for_catchup() here; it isn't sufficient
# when tablesync workers might still be running. So in addition to that,
# we verify that no tablesync workers appear for the subscription.
# XXX maybe this should be integrated in wait_for_catchup() itself.
$node_twoways->wait_for_catchup('testsub');
$node_twoways->poll_query_until(
'd2',
"SELECT count(*) FROM pg_stat_subscription WHERE subname = 'testsub' AND relid <> 0",
"0");
is($node_twoways->safe_psql('d2', "SELECT count(f) FROM t"),
$rows * 2, "2x$rows rows in t");
is($node_twoways->safe_psql('d2', "SELECT count(f) FROM t2"),
$rows * 2, "2x$rows rows in t2");