Crash in 5.5.15-55

[Dan Rogers]

We have a Percona Server 5.5.15-55 install with a master (db01) and two slaves (db03 & db04), running replication between the servers. db03 is also the master for db01 for cases where we need to bring db01 down for maintenance.

Yesterday, we we had a near-simultaneous crash of MySQL on all three servers, leading me to believe it was caused by a query that propagated between them.

I've attached the relevant log entries from the MySQL log. Please let me know if I can provide any further information.

Thanks.

Dan.

[Alexey Kopytov (akopytov)]

The crash is a result of trying to create a temporary InnoDB table on tmpfs. The reason is that async I/O is not supported on the tmpfs filesystem, see http://bugs.mysql.com/bug.php?id=58421

A workaround is to either set innodb_use_native_aio=0 or point tmpdir to an on-disk filesystem.

[Dan Rogers (drogers-l)]

Thank you for the quick reply.

I've pointed tmpdir back to a physical disk to avoid this problem. You may close the bug if you wish.

Dan.

[Alexey Kopytov (akopytov)]

The upstream bug has been fixed in MySQL 5.5.21 and 5.6.5. Setting to FIx Released in PS 5.5.

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PS-503