Questions about mupdate setup and murder in general

[Lenny]

I'm running a cyrus murder setup (with virtual domain support) on version 2.2.3
(Simon's RPMs). There's about a dozen domains with between 26,000 - 27,000
users running on 2 frontends (soon to be 3 dell poweredge 1750's) and 1 backend
(dell poweredge 2650).

I just moved our last domain onto the cyrus system. Approximately 16,800 users.
Migration went fine until I pointed incoming mail delivery at the frontends and
then everything fubar'd. The problem is, I can't seem to figure out what the
issue is. The symptoms are as follows:

1) The master mupdate server (a seperate machine from the frontends and backend)
will start throwing errors like this: Jul  9 10:41:32 db mupdate[2515]:
badlogin: cyr-fe2.xxxxx [xxxxxxxx] DIGEST-MD5 SASL(-13): authentication
failure: required parameters missing
This only happens when everything else starts going crazy.

2) sendmail on the frontends will start returning various error messages.
Sometimes just "Deferred", sometimes simply "queued". But mostly I get messages
like these:
Jul  9 10:46:13 cyr-fe1 sendmail[362]: i69Ejxb4000334: to=<dtdr at xxxx>,
delay=00:00:08, xdelay=00:00:06, mailer=cyrusv2d, pri=88656, relay=localhost,
dsn=4.4.2, stat=Deferred: Connection reset by localhost
I can't seem to find any debug information (via strace, etc) to tell me WHY the
connection is being reset. Once these start showing up in the logs, I know
everything's gone to hell in a handbasket. I have to then stop sendmail on both
frontends, saslauthd and cyrus-imapd and restart them all to get everything
working again. Pop and imap logins stop working also.

3) The next step is that it complains about not being able to contact the
mupdate server. The mupdate server appears to be totally fine tho.

4) Oddly enough, when I shut down cyrus-imapd on the backend, the frontends
complain about not being able to contact the mupdate server (although this is
pointed at a seperate server in imapd.conf). It complains about this INSTEAD of
"can't contact backend".

Could it be that the sheer number of connections to 1 backend is causing it to
get hosed up? Or the same, but with the frontends? Mupdate server can't keep
up? I'm so stumped. As you can imagine this is causing a major problem for me
:)

Any other suggestions for debug?

Load on the boxes themselves is low (1-2 load average, they're all dual
processor boxes, fedora core 1).

Thanks.

Lenny
--

"Wisdom is to a man an infinite Treasure" - Anonymous"



---
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