70k mqueue.in but load under 1 ??

Furnish, Trever G TGFurnish at herffjones.com
Wed Nov 29 15:58:02 GMT 2006


> -----Original Message-----
> From: mailscanner-bounces at lists.mailscanner.info 
> [mailto:mailscanner-bounces at lists.mailscanner.info] On Behalf 
> Of Richard Frovarp
> Sent: Tuesday, November 28, 2006 11:55 AM
> To: MailScanner discussion
> Subject: Re: 70k mqueue.in but load under 1 ??

> If the queue is still building up, something is most likely 
> timing out. 
> Are you running Pyzor, Razor, or DCC? Using any odd URIBLs or 
> DNSBLs that could be timing out? Several months ago Pyzor 
> timed out (10 seconds per message) on me for a while, that 
> really hurt performance. Do a lint or debug and see if 
> anything hangs for a noticeable time period as Glenn suggested.

Just some random follow-up thoughts.

My own problem became less urgent when I went ahead and put SBL+XBL in
at the MTA level, which has effectively cut our inbound message count in
half.  Pleasant surprise there.

I looked for things timing out, but I didn't get far enough to get exact
timing information for the spamassassin lint output.  I wish I had been
keeping a history of how long the razor and bayes checks take on my
system on a single test message every day.  I'm adding that to my to-do
list so that I have a baseline in the future.

I expected to find an RBL timing out or spamassassin timing out, but I
didn't find that.  I did note that for most of the morning on the day I
had the biggest problem our internet pipe was maxed out, so DNS checks
may have been slower, which would of course slow SA more than the MTA
(cuz I only do a couple of dns checks in the MTA, many in SA).

In the past I've gotten a boost in performance by starting with a fresh
Bayes db after the db had grown very large.  Haven't done that in a
while, might be part of the slowdown.

More information about the MailScanner mailing list