Affects Version/s: None
Fix Version/s: None
**Reported in Launchpad by Jeremy Tinley last update 13-05-2015 05:03:29
Am able to break multi-threaded slave (SQL thread) in Percona Server 5.6.22-71 with a 1755 error after restarting the server a few times.
24 schemas (db1, db2.. etc to db24)
Simple bash script inserting an auto-increment value randomly into a different database.
Replica has the following:
master_info_repository = TABLE
relay_log_info_repository = TABLE
slave_parallel_workers = 16
While running ~100 copies of the attached script to generate insert traffic, I would restart the replica. This will eventually lead to an error similar to the following:
2015-02-11 00:09:17 93614 [Note] Slave I/O thread: connected to master 'email@example.com:3306',replication started in log 'mts1-bin.000001' at position 495366437
2015-02-11 00:09:17 93614 [Note] Event Scheduler: Loaded 0 events
2015-02-11 00:09:17 93614 [Note] /usr/sbin/mysqld: ready for connections.
Version: '5.6.22-71.0-log' socket: '/var/lib/mysql/mts2/mysql.sock' port: 3307 Percona Server (GPL), Release 71.0, Revision 726
2015-02-11 00:09:17 93614 [Note] Slave SQL thread initialized, starting replication in log 'mts1-bin.000001' at position 369721635, relay log '/var/lib/mysql/mts2/mts2-relay.000284' position: 55398130
2015-02-11 00:09:17 7f524f5f7700 InnoDB: Buffer pool(s) load completed at 150211 0:09:17
2015-02-11 00:09:19 93614 [Warning] Slave SQL: injecting ROLLBACK at the end of cold group, Error_code: 0
2015-02-11 00:09:19 93614 [ERROR] Slave SQL: Cannot execute the current event group in the parallel mode. Encountered event Xid, relay-log name /var/lib/mysql/mts2/mts2-relay.000286, position 435 which prevents execution of this event group in parallel mode. Reason: the event is a part of a group that is unsupported in the parallel execution mode. Error_code: 1755
I'm guessing that this merge: https://code.launchpad.net/~sergei.glushchenko/percona-server/5.6-ST43091-ps-bug1331586/+merge/229168 does not handle MTS correctly. I reproduced it a few times and every time the "injecting ROLLBACK at the end of cold group" message appeared prior to every breakage. The event that breaks is always the first event in the relay log.
I was unable to reproduce this in MYSQL Community Edition 5.6.22.
Attached is 'generate_insert_load.sh', the script I use to generate the inserts on the master. I ran 100 copies to generate inserts.