Termination of the session is omitted sporadically from the slow query log

[Laurynas Biveinis]

I don't have enough information for a full reproducible procedure, so just opening this for a possible follow-up in the future.

Usually session termination is slow-logged as "# Administrative command: Quit", if slow log options allow it. However, occasionally it is not logged, with no clear pattern.

For example, http://jenkins.percona.com/job/percona-server-5.1-param/498/BUILD_TYPE=debug,Host=debian6-64/testReport/junit/%28root%29/main/percona_mysqldumpslow/ shows it to be logged 125 times out of 128.

This is with
log_slow_admin_statements=ON;
long_query_time=0;
min_examined_row_limit=0;

Currently not sure about upstream and 5.5.

[Laurynas Biveinis (laurynas-biveinis)]

Setting triage to Low as it's unlikely to affect users in any important way.

[Stewart Smith (stewart)]

Laurynas Biveinis <email address hidden> writes:
> I don't have enough information for a full reproducible procedure, so
> just opening this for a possible follow-up in the future.
>
> Usually session termination is slow-logged as "# Administrative command:
> Quit", if slow log options allow it. However, occasionally it is not
> logged, with no clear pattern.

This may affect people using percona-playback to play back captured
load, leaving the replay with a lot of hanging connections not doing anything.

--
Stewart Smith

[Laurynas Biveinis (laurynas-biveinis)]

I haven't considered using slow query logs as an input to something else, with Quit as a session termination marker. That could be more serious, resetting triage then.

[Stewart Smith (stewart)]

Laurynas Biveinis <email address hidden> writes:
> I haven't considered using slow query logs as an input to something
> else, with Quit as a session termination marker. That could be more
> serious, resetting triage then.

I'll set to Medium as is probably higher up than most of our Low ones,
and we certainly have a bunch more higher prio bugs too.

--
Stewart Smith

[Laurynas Biveinis (laurynas-biveinis)]

See also bug 1163680.

[Yura Sorokin (yura-sorokin)]

No signs of this test failure on both 5.5 and 5.6 in Jenkins for at least 1 months (as of September 21, 2015).
Setting status to "Incomplete".

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PS-1295