A client of mine runs MySQL in a pretty high volume environment, and for
the past couple months we've had a lot of trouble with corrupt tables
(error 127). We've changed the MySQL version (presently to version
3.23.49), FreeBSD version (presently to 4.5-STABLE), hardware (from IDE
to SCSI) and of course the amount of data involved. We aren't using
LinuxThreads, soft updates, async filesystems, or any other edgey
performance options. The system was built by Rackspace so I can assume
decent quality components, and besides the problem has followed us after
a server change.
The error occurs almost at random. Usually once or twice a week, always
on the same table, and not necessarily at high load times.
The table involved (so far it has only happened on one table) is
frequently truncated, and it has an almost 1:1 ratio of selects and
inserts.
I can always use "myisamchk", but usually this ends up deleting every
record in the table. (Luckily for us this is an okay situation based on
our application) Running myisamchk on a scheduled basis is a bit too
much of a hack for me to explain to my client -- and my conscience.
What can I do to diagnose and resolve this problem?
--
tlack@stripped
