I had a similar problem were an apache process would starting up all of
the cycles on the CPU it was running on. If I did not restart apache
every day Apache would stop serving RT requests while using %100 of the
cycles on all four processors.
I fixed the problem by moving from Apache 2.0.48 using the worker MPM
(The default for Debian Stable Apache2 installs) to using the prefork
MPM. The problem went away and I have not seen it reoccur in over a
month. I was also experiencing incredibly slow start times when
launching apache initially and now it still takes a little while but not
as bad as before.
I recall reading that one of the CPAN packages that RT requires is not
thread safe, so I assume that that is what the problem was. This did not
ever occur on my testing box, so I am assuming it is related to
something that one of my users does on a frequent basis, possibly
related save searches, but I am just guessing. I am also not (%100) sure
that this was RT library problem it might be related to mod_kerberos,
since I am also using that.
Student Affairs Information TechnologyFrom: Jesse Vincent [mailto:firstname.lastname@example.org]
Sent: Thursday, December 07, 2006 7:56 AM
To: Robert Grasso
Subject: Re: [rt-users] rt 3.6.1 : apache load sometimes climbs to 100%
: it’sRT fault ?
I just discovered
and well : it’s just as slow as mine, exactly in the SAME way. When
you open the first ticket after
having performed a search, it takes 20 seconds to display the page -
further pages take nearly 1
second to open.
I really should get around to updating that RT to something released
within the past year. (Also, for better or for worse, the box running
rt3.fsck.com is also running our Kwiki and our spam filtering. Both of
which are…heavy users of CPU and disk:
07:54:41 up 101 days, 15:00, 16 users, load average: 12.28, 9.07, 6.91
Also, please don’t cross-post to both mailing lists. generally, topics
really belong on one or the other.