Limiting SpamAssassin Children?

[jshanley]

Since Plesk still isn't using SpamAssassin v3.x (2.63 is the version that ships with Plesk), I don't see an easy way to limit the maximum number of spamc children. Thus, when one of the servers gets spammed heavily, the server spawns off 30-100 copies of SpamC - sucking up all the CPU and memory on the system.

Has anyone limited the spamc processes successfully? The "-m" switch doesn't seem to work on 2.63.

-J

 

[graffix]

Hi,

i have the same problem:

Getting a high load of spam, that crash the server down. Because the memory is running out and the swap too.
So is there a help for this?

 

[zeki79]

i have the same problem

 

[graffix]

And the best story is:

We have server from dell, and the machines stucks in a few days.
First SMTP is crashing down. The second is httpd is stopped working.
After this, the server pings but no services availble. In this case we must make a hard reset.

That is really NOT good.

Our Confixx Servers are running fine over months ago.
So why is plesk so risky to use?

Our next step is to move all data to Virtuozzo and hope that the system is stable after that.

 

[jshanley]

Limiting SPAMD children

I have heard that adding:

SPAMASSASSIN_MAX_CHILDREN 10

... to /etc/psa.conf may limit the number of children with Plesk. I'll go test it and see if that works out.

 

[Whistler]

Originally posted by graffix
Our Confixx Servers are running fine over months ago. So why is plesk so risky to use?

I've used Plesk for 1½ year now - and have only had about 2 x 30 minutes downtime (besides a couple of upgrades that lasted a couple of minutes when all was prepared and some minor network dropouts) - 1 of the 2 times was when I had to physically change datacenter - the second time was a broken uplink out of the datacenter.

So I would say that Plesk is not risky to use - but it's risky to use any software/server not closed completely down by a firewall if you don't know what your'e doing!

I've said it before and I'll keep repeating it - Plesk can't be put on a server to replace a skilled (expensive) technician/server admin. Plesk is not a computerized administrator that'll do all for you.

Plesk is a tool to help a server administrator doing certain triviel (and sometimes non triviel) management - and is put on top of your server helping your customers with a nice interface to do their daily job (saving you the time of doing it for them).

But Plesk can NEVER replace a server administrator! No matter what your'e told, promised or "sold".

(just to clear that out - sorry) ;o)

 

[jonathanjab]

Re: Limiting SPAMD children

Originally posted by jshanley
I have heard that adding:

SPAMASSASSIN_MAX_CHILDREN 3

... to /etc/psa.conf may limit the number of children with Plesk. I'll go test it and see if that works out.

Yep - that worked a treat! Although the file is /etc/psa/psa.conf on our servers!

 

[atomicturtle]

side note, what that means is that you'll only run 3 spamd children. The effect is that you've limited spamassassin for all your mailboxes. IE, if spamc takes more than 30 seconds to process a message, its just going to let it through. Thus eliminating spamassassin filtering for most of your users.

Try throttling your inbound smtp connections instead.

 

[jonathanjab]

That's fine - the throughput on the server is no-where near enough to max out SpamAssassin with 3 children (one e-mail every 2-5 minutes on average).

The issue for us is better memory management on the server is question to reduce the risk of overloading the server. SpamAssassin is very inefficient at sharing memory and settings and duplicates alot of details across all the children.

 

[atomicturtle]

If a handful of spamassassin processes are killing your system now, then youve got other issues. IE, slow dns lookups, or youve been blacklisted by one of the signature services/rbls, bad cpan installs, etc. I'm handling around 30 concurrent spamc threads at all times on my systems, and its not killing a low end P4 with a gig of ram.

 

[jonathanjab]

SpamAssassin isn't the problem. It's all part of a general tune-up of many of the services on the server to improve it's memory usage together while dealing with a number of sites that have issues with some of their scripts using excessive memory on the server.