2012/10/4 Andrew Miklas <and...@pagerduty.com> > Hi guys, > > I recently had a data corruption issue with InnoDB. MySQL was shut down > improperly (power failure), and when the system came back up, MySQL refused > to start. On inspection of the logs (see below), it looks like the > tablespace became seriously corrupted. In the end, I had to rebuild the > slave using mysqldump. > > I'm curious what happened here, since I thought InnoDB wasn't supposed to > become corrupted on an improper shutdown. One possibility that we were > exploring was that the filesystem journal setting was incorrect. We were > using ext3 with the journal set to writeback mode. Is this a known bad > config with InnoDB? >
Hey Andrew, it shouldn't be a biggie if you have a BBU. Do you guys use HW RAID + BBU? What's your innodb_flush_log_at_trx_commit value? Have you tried playing with innodb_force_recovery option to try to get the server started at least? That way you might be able to identify which table(s) is/are the corrupted one and the one(s) preventing the whole server from booting up. Manuel
