Bug #34718 mysqld crash on resume
Submitted: 21 Feb 2008 9:59 Modified: 25 Feb 2009 9:21
Reporter: Tibor Horvath Email Updates:
Status: Won't fix Impact on me:
None 
Category:MySQL Cluster: Cluster (NDB) storage engine Severity:S2 (Serious)
Version:5.0.45-Debian_1-log OS:Linux (Debian etch, Linux 2.6.22-3-486)
Assigned to: Magnus Blåudd CPU Architecture:Any

[21 Feb 2008 9:59] Tibor Horvath
Description:
Mysqld gets signal 11 after OS resumes from hibernation.

MySQL is used with JBoss 3.2.7 (Connector/J 3.1.6) running on the same machine. Mysqld is an SQL node for a MySQL Cluster. It has 1 local MyISAM table and the rest of the database is in Cluster.

Shortly after resuming the OS, JBoss starts processing requests, and then mysqld dies. The actual query being executed differs across crashes, but the stack trace is identical. See attached log and resolved stack trace.

How to repeat:
Start MySQL Cluster, suspend machine, resume machine, immediately start issuing queries that access cluster tables.
[21 Feb 2008 10:02] Tibor Horvath
syslog

Attachment: mysqld.bug (application/octet-stream, text), 8.64 KiB.

[21 Feb 2008 10:03] Tibor Horvath
resolved stack trace

Attachment: mysqld-resolved.stack (application/octet-stream, text), 748 bytes.

[21 Feb 2008 11:08] Tibor Horvath
Sorry, the Connector/J version is 5.0.4. (Probably doesn't make a difference.)
[25 Feb 2009 9:21] Magnus Blåudd
Problem seems to occur when mysqld has lost connection to cluster and reconnects again. Workaround would be to kill and restart the mysqld when resuming again.

Problem have been fixed by patches for BUG#33626 in MySQL Cluster 6.3.19