Seems a little underpowered to me.
10 children w/ 448M is pushing it, IMHO.
Same opinion of a 550 P3 scanning EMail with
SA and clamAV.
Given your system, I would:
1) Reduce max-children to 8 or less. (Use 6 if you can).
2) Eliminate pre-forking (--round-robin option).
3) Set a timeout option (--timeout-child=60).
4) Reduce message size passed to SpamAssassin to
32K or less. There is a cost/benefit ratio to
consider here. What percentage of your spam is
greater than 32K?
5) Reduce stack size (ulimit) for the spamd process
(if you can) to improve memory efficiency.
6) Pray that some botnet doesn't latch on to you
and obliterate your server!
But that's me. I am assuming your system had a
working configuration with a stock SA installation.
Good luck.
Jean-Paul Natola wrote:
1.We have a sitewide config
2.Startup parameters are
command_args="-d --min-children=2 --max-children=10 --min-spare=2
--max-spare=3 --max-conn-per-child=100 -r ${pidfile}"
3.I do not scan messages over 250 k
4.We process roughly 5k messages daily
5. the hardware is ehh kinda crappy, but we don't really process that much
PIII 550 448 ram
And this machine does not "hold" any mail , it process it, then , if clean
and spam free it goes right to our exchange server
-----Original Message-----
From: Daryl C. W. O'Shea [mailto:[EMAIL PROTECTED]
Sent: Thursday, March 13, 2008 8:44 PM
Cc: SpamAssassin
Subject: Re: Plugin eval failed
On 13/03/2008 8:02 PM, Jean-Paul Natola wrote:
What other info should I post?
The output of vmstat while this is happening.
A definition of "constantly". Every message, some % of messages? For
some common size of messages? Whether it happens when scanning mail for
a subset of users or for all users. Relevant spamd log output.
spamd debug output.
Your spamd startup parameters along with a description of the hardware
you're running.
A summary of any non-default rulesets you're using.
A comment regarding the validity, in this instance, of my comment about
the common causes of the child processing timeout symptom.
Daryl
