[Mimedefang] High CPU Utilization

[David F. Skoll]

On Wed, 5 Mar 2003, Michael Pellegrino wrote:

> Pentium II, 256MB Ram, MD 2.30, SA 2.50, clamd 2.54, SPOOLDIR resides on a
> RAM Disk, server does maybe 1000-1200 messages per day tops.  10 min slaves,
> 30 max.  timeout 90, busy timeout 300.  Suse Linux 8.1 (2.4.19).

What's the clock frequency on your box?

10 min slaves is way too high for that message volume; I'd recommend
2 or 3 slaves.

> When sendmail, md, and md-mx are all starting, I get a bunch of
> mimedefang.pl processes starting up which is perfectly normal.

You can use MX_SLAVE_DELAY (or mimedefang-multiplexor's "-w" option) to
stagger the initial starting times of the slaves.

> However,
> they all show anywhere between 18-34% CPU utilization.  This lasts for about
> a minute or so until they all drop to 0.

Thank you, Perl. :-)  That's Perl's initialization.

> Also, they will jump up to that
> range for each email message that it processes.

Well, I would expect close to 100% CPU utilization while a message is
being processed, but it shouldn't last very long -- a few hundred
milliseconds in most cases.

> We occasionally have brief spikes in email traffic, about 20-30 messages
> over a 2-4 minute time period.

That's not a lot, and your server should handle it.  Decrease the
min_slaves; you may be swapping with 10 slaves.

--
David.