Matt Kettler wrote:
jdow wrote:
From: "Derek Catanzaro" <[EMAIL PROTECTED]>
Matt Kettler wrote:
Derek Catanzaro wrote:
I have been having issues with mail backing up on and off over the
past week. I am using MailScanner with SpamAssassin. This morning
for example, I had roughly 500 messages waiting in
/var/spool/mqueue.in and that number had increased to about 2200 in
less than an hour. I then tell MailScanner to stop using SpamAssassin
to try and identify if the problem is with SpamAssassin or not and now
I'm back down to less than 50 messages waiting in the queue in less
than a matter of 10 -15 minutes. So obviously this tells me something
is going on with SpamAssassin.
I ran "spamassassin --lint -D" and I did not notice any problems with
the output other than a dcc timeout. Then again, spamassassin has
always worked well for me so I may be missing something in the output
because I have really never had to troubleshoot this kind of issue
with spamassassin. The recent changes I have made to try and combat
the problem is to disable bayes and I turned off the auto expire for
the bayes tokens just to make sure that wasn't slowing things down.
I am running a local caching name server so I do not believe this to
be a DNS timing issue. I can provide my spamassassin --lint -D output
if anyone is interested.
Fedora Core 1
SpamAssassin 3.1.0
MailScanner 4.49.7
sendmail 8.13.5
Thanks,
Derek
What's your memory load look like? (ie: run the "free" command).
Have you recently added any add-on rulesets?
Do you have a whole pile of bayes_toks files suffixed with a process ID
and "expire" laying around in your bayes directory?
Here are the results of the "free" command with spamassassin running:
total used free shared buffers cached
Mem: 2068504 2041572 26932 0 242712
60556
-/+ buffers/cache: 1738304 330200
Swap: 1831912 58544 1773368
Results of "free" command without spamassassin running:
free
total used free shared buffers cached
Mem: 2068504 1712204 356300 0 244080
73944
-/+ buffers/cache: 1394180 674324
Swap: 1831912 7172 1824740
Subtract at least 1 from the number of children you allow for
spamassassin if you can. (I don't know how mailscanner works.)
Going into swap with SpamAssassin is pure poison.
I'd have to agree.. either that or move SA, or some other part of that
box's load off somewhere else.
I'd generally consider the numbers you're posting for the box without
SA as running as being a "healthy but fully loaded" server.
Thanks for the suggestions. I will try reducing the number of
children. The issue that was caused yesterday was due do dcc timeouts.
I disabled the dcc checks and mail was routing in a timely manner, the
backup went away. This morning I'm stuck with the same thing again, but
now pyzor and dcc are timing out. These inconsistencies are really
nerve racking. I have had this system running for a couple of years now
and have not run into these problems and all of a sudden within the last
week this occurs.
I have checked with my WAN group and no firewall rules have been
changed. They are allowing the ports for pyzor, razor, and dcc (as well
as DNS and SMTP) so I'm at a loss.... If you folks experience timout
issues with dcc or pyzor does it cause a backup with your mail or am I
the only one (I don't think I would be)?
Thanks,
Derek
--
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.
