database/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2025-07-29 05:21:33 +03:00
Code Activity

MDEV-6775: Wrong binlog order in parallel replication: Intermediate commit

Browse Source 
The code in binlog group commit around wait_for_commit that controls commit
order, did the wakeup of subsequent commits early, as soon as a following
transaction is put into the group commit queue, but before any such commit has
actually taken place. This causes problems with too early wakeup of
transactions that need to wait for prior to commit, but do not take part in
the binlog group commit for one reason or the other.

This patch solves the problem, by moving the wakeup to happen only after the
binlog group commit is completed.

This requires a new solution to ensure that transactions that arrive later
than the leader are still able to participate in group commit. This patch
introduces a flag wait_for_commit::commit_started. When this is set, a waiter
can queue up itself in the group commit queue.

This way, effectively the wait_for_prior_commit() is skipped only for
transactions that participate in group commit, so that skipping the wait is
safe. Other transactions still wait as needed for correctness.
This commit is contained in:
Kristian Nielsen
2014-11-13 10:31:20 +01:00
parent eec04fb4f6
commit d08b893b39
5 changed files with 250 additions and 136 deletions
Download Patch File Download Diff File Expand all files Collapse all files

69
mysql-test/suite/rpl/t/rpl_parallel.test
View File

@ -1466,6 +1466,75 @@ SET sql_slave_skip_counter= 1;
SELECT * FROM t2 WHERE a >= 30 ORDER BY a;
--echo *** MDEV-6775: Wrong binlog order in parallel replication ***
--connection server_1
# A bit tricky bug to reproduce. On the master, we binlog in statement-mode
# two transactions, an UPDATE followed by a DELETE. On the slave, we replicate
# with binlog-mode set to ROW, which means the DELETE, which modifies no rows,
# is not binlogged. Then we inject a wait in the group commit code on the
# slave, shortly before the actual commit of the UPDATE. The bug was that the
# DELETE could wake up from wait_for_prior_commit() before the commit of the
# UPDATE. So the test could see the slave position updated to after DELETE,
# while the UPDATE was still not visible.
DELETE FROM t4;
INSERT INTO t4 VALUES (1,NULL), (3,NULL), (4,4), (5, NULL), (6, 6);
--source include/save_master_gtid.inc
--connection server_2
--source include/sync_with_master_gtid.inc
--source include/stop_slave.inc
SET @old_dbug= @@GLOBAL.debug_dbug;
SET GLOBAL debug_dbug="+d,inject_binlog_commit_before_get_LOCK_log";
SET @old_format=@@GLOBAL.binlog_format;
SET GLOBAL binlog_format=ROW;
# Re-spawn the worker threads to be sure they pick up the new binlog format
SET GLOBAL slave_parallel_threads=0;
SET GLOBAL slave_parallel_threads=10;
--connection con1
SET @old_format= @@binlog_format;
SET binlog_format= statement;
SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued1 WAIT_FOR master_cont1';
send UPDATE t4 SET b=NULL WHERE a=6;
--connection server_1
SET debug_sync='now WAIT_FOR master_queued1';
--connection con2
SET @old_format= @@binlog_format;
SET binlog_format= statement;
SET debug_sync='commit_after_release_LOCK_prepare_ordered SIGNAL master_queued2';
send DELETE FROM t4 WHERE b <= 3;
--connection server_1
SET debug_sync='now WAIT_FOR master_queued2';
SET debug_sync='now SIGNAL master_cont1';
--connection con1
REAP;
SET binlog_format= @old_format;
--connection con2
REAP;
SET binlog_format= @old_format;
SET debug_sync='RESET';
--save_master_pos
SELECT * FROM t4 ORDER BY a;
--connection server_2
--source include/start_slave.inc
SET debug_sync= 'now WAIT_FOR waiting';
--sync_with_master
SELECT * FROM t4 ORDER BY a;
SET debug_sync= 'now SIGNAL cont';
# Re-spawn the worker threads to remove any DBUG injections or DEBUG_SYNC.
--source include/stop_slave.inc
SET GLOBAL debug_dbug=@old_dbug;
SET GLOBAL binlog_format= @old_format;
SET GLOBAL slave_parallel_threads=0;
SET GLOBAL slave_parallel_threads=10;
--source include/start_slave.inc
--connection server_2
--source include/stop_slave.inc
SET GLOBAL slave_parallel_threads=@old_parallel_threads;