Looks like we had
https://bugzilla.andrew.cmu.edu/show_bug.cgi?id=2677
which has been fixed in a newer release then what we were running.
Basically, mupdate would create to many threads (your max threads) then
kill them until you had (max idle threads). This was happening over and
over for us, until some system resource was reached and then
pthread_create would fail.
Upgraded to the most recent release and the problem has not re-occured.
-Patrick
On Tue, 24 Jan 2006, Patrick Radtke wrote:
We've been seeing a lot of mupdate worker thread problems the last 2 days.
Things like
Jan 24 12:38:42 notdog mupdate[5295]: could not start a new worker thread
(not fatal)
Jan 24 12:38:42 notdog last message repeated 353 times
Jan 24 12:38:45 notdog mupdate[5295]: login: mettwurst.cc.columbia.edu
[128.59.33.138] backend PLAIN User logged in
we're trying to debug the cause of the problem, but I'd thought I'd ask on
the list if anyone had any insights.
Logins are infrequent and generally there aren't that many connections
established.
lsof | grep "mupdate->" | wc -l
18
We increased the max number of worker threads, and max connections
mupdate_workers_max: 400
mupdate_connections_max: 1792
We increased a lot of the ulimits
and increase the maxfds
mupdate cmd="mupdate -m" listen="mupdate" prefork=1
maxfds=5048
We're running on RHEL4 and Cyrus 2.3
Thanks for any help or pointers.
-Patrick
----
Cyrus Home Page: http://asg.web.cmu.edu/cyrus
Cyrus Wiki/FAQ: http://cyruswiki.andrew.cmu.edu
List Archives/Info: http://asg.web.cmu.edu/cyrus/mailing-list.html
----
Cyrus Home Page: http://asg.web.cmu.edu/cyrus
Cyrus Wiki/FAQ: http://cyruswiki.andrew.cmu.edu
List Archives/Info: http://asg.web.cmu.edu/cyrus/mailing-list.html
