Bug #30461 Mysqld signal 6 when tmpdir fills
Submitted: 16 Aug 2007 16:55 Modified: 24 Aug 2007 20:32
Reporter: Narayan Newton Email Updates:
Status: Closed Impact on me:
None 
Category:MySQL Server: General Severity:S3 (Non-critical)
Version:5.0.40 OS:Any (x86_64)
Assigned to: CPU Architecture:Any
Tags: crash

[16 Aug 2007 16:55] Narayan Newton 
Description:
I have a very odd problem that keeps coming up roughly every 24 hours. I have two servers in a master-slave setup and the queries are split between them. Because of how this is setup, the slave gets most of the queries that require on disk temp tables. There are so many of these, that we have tmpdir pointed at a tmpfs mount. This works fine, except every once in awhile the slave crashes with a signal 6 from glibcs malloc checks. The errors always occur right after tmpdir fills up and we get a bunch of 'Incorrect key file for table' errors for the temp tables. Sometimes it recovers from them and sometimes it crashes. I have not been able to reproduce this by manually filling up tmpdir, but its happened on two servers now so I doubt its hardware. Attached is the my.cnf and the most "enlightening" error log (the one with a backtrace).

How to repeat:
Can't Reproduce
[16 Aug 2007 16:55] Narayan Newton 
Error log

Attachment: mysqld.err (application/octet-stream, text), 123.29 KiB.
[16 Aug 2007 16:57] Narayan Newton 
my.cnf

Attachment: my.cnf (application/octet-stream, text), 3.22 KiB.
[16 Aug 2007 23:58] Narayan Newton 
Pretty sure this is #28449, I'll close this when I confirm it.
[24 Aug 2007 20:32] Narayan Newton 
Confirmed as a duplicate, sorry all.