server crashes ocasionally on replication

[Dima]

Recently upgraded to Percona 5.5.20. Slave crashes 2-3 times a day or hangs and needs to be killed.

When it crashes it is always in this stack:
*** glibc detected *** /usr/sbin/mysqld: double free or corruption (!prev): 0x00007f2a14378420 ***
======= Backtrace: =========
/lib64/libc.so.6[0x32b8476d63]
/usr/sbin/mysqld(_ZN15Query_log_eventD0Ev+0x28)[0x56b918]
/usr/sbin/mysqld[0x52e2c5]
/usr/sbin/mysqld(handle_slave_sql+0x919)[0x52f4a9]
/lib64/libpthread.so.0[0x32b8806ccb]
/lib64/libc.so.6(clone+0x6d)[0x32b84e0c2d]

When it hangs there is always this message:
*** glibc detected *** /usr/sbin/mysqld: malloc(): memory corruption: 0x00007fe418187040 ***

There is a ton of other info in the log. Let me know what else is needed. This is a heavily replicating slave. I don't think I know how to reproduce it or what causes it but it happens very often.

[Sheeri K. Cabral (awfief)]

I'm also getting this crash on 5.5.20:

mysql> show variables like 'version';
+---------------+---------------+
| Variable_name | Value |
+---------------+---------------+
| version | 5.5.20-55-log |
+---------------+---------------+
1 row in set (0.00 sec)

Unlike https://bugs.launchpad.net/percona-server/+bug/705688, the query_cache_strip_comments is off:
mysql> show variables like 'query_cache%';
+------------------------------+-----------+
| Variable_name | Value |
+------------------------------+-----------+
| query_cache_limit | 1048576 |
| query_cache_min_res_unit | 4096 |
| query_cache_size | 268435456 |
| query_cache_strip_comments | OFF |
| query_cache_type | ON |
| query_cache_wlock_invalidate | OFF |
+------------------------------+-----------+
6 rows in set (0.00 sec)

I have just set query_cache_size=0 and query_cache_type=0, I'm hoping that will fix this problem.

[Alexey Kopytov (akopytov)]

It's most likely a duplicate of bug #915814. Feel free to reopen if the problem persists after a fix for that bug is available.