Spamassassin timeouts - Just an observation
campbell at cnpapers.com
Sun Jan 18 11:32:03 GMT 2009
Quoting "Koopmann, Jan-Peter" <jan-peter at koopmann.eu>:
> > The topic seems to come up quite often, and although the answers are
> > usually pretty much the same, I never really see much of a "Solved"
> are you using BotNet.pm by any chance. There was a bug in one of the
> older versions causing sporadic SpamAssassin timouts.. I looked for ages
> and on my systems the old BotNet.pm triggered it. Updated (without
> changing anything else) and never seen the error again. Just an indea.
So far, it appears the extra rules from SARE was the biggest contributor. I have
removed all of the sets from my sa-update and the problems almost disappeared. I
do not run BotNet.pm.
The most common problem with these timeouts always seemed to be DNS and RBLs,
but I wasn't seeing any problems there. I kept looking there though. I was also
being fooled by high, but not critical load averages. I have duplicate servers
that were not timing out with similar load averages, rules, and daily email
counts. The non-problem machine was getting it's email spread out over the
course of a day, whereas the problem machine was receiving large batches at
different times of the day.
Once I started reviewing the mailscanner-mrtg plots, I saw this. Another thing
that threw me off was the fact that no matter how many emails arrived at one
time, the LA would spike to 3.5 or higher on either machine. The high message
per batch count would cause the LA to gradually creep higher, but the smaller
batches would give constant LAs. The low amount of RAM for both machines explain
I had been fooled by MS doing such a good job for years, and just wasn't
thinking very clearly about what could have caused this. Two upgrades ago, I
started using the new sa-update feature and added the rules using that. It
didn't show immediate changes to the way the machines acted over a week or so,
so I never thought it was a problem. The load averages are still fluctuating,
but batch times are considerably lower, which allows faster throughput, and less
timeouts on the machines. I hope the RAM I have ordered will fix the rest of it.
I'm sorry to have caused such a stir with all of this, as this thread has went
on way to long. I've sharpened my MS diagnostic skills, though, and hope it
might have helped others - the information everyone has provided has been very good.
Thanks to all again,
This mail sent through IMP: http://horde.org/imp/
More information about the MailScanner