Qmail on RH 6.2 problems

[paleogryph]

Qmail running on RH 6.2, p3 600 with 1gb ram.

Users complain of slow performance, check the server. Find an error message scrolling on each terminal, disallowing any login.
The error, "try_to_free_pages" wil scroll and displays the name of each daemon running after it.

Reboot the server.

When server comes up, it functions again but after about and hour or two the problem comes back. This is after this box has been working fine for months. Once system is back up, I run df and see there is lots of free space and no unusual processes running. Cat /proc/meminfo shows that the system is real close to using all the memory. Also see that /var/log/maillog hasn't been updating itself as per /etc/syslog.conf. The file is over a day old and its updating very slowly.

Any ideas?

 

[sleipnir214]

Make sure that your box isn't being used as an open mail relay, for one. Want the best answers? Ask the best questions:

http://www.tuxedo.org/~esr/faqs/smart-questions.html

TANSTAAFL!

 

[marsd]

What kernel version, and are there any other memory or
"flaky" type messages?

If it just happened as you say chances are either you
got cracked and something else is going on that you are
not seeing, or your memory and/or processor or possibly
other hardware is going(gone)bad...

I used to have random memory/paging faults
after extended uptimes with an older,first generation
athlon box that was underventilated.
After a week or two processes would die and assorted weirdness occurred.
I had a similar problem with a dns server that
had developed bad sectors on the hd.

Usually if it's out of the blue and you feel like you
are up to date on your security, hardware is a
good place to start looking.

 

[paleogryph]

Ok, I found out what the problem was.

We run Qmail on this box, which is where the actual user mail accounts live. We also run SpamAssassin v 2.31 on here. SpamAssassin will connect to munitions2.xs4all (which is Vipuls Razor, a "collaborative" spam testing repository) to get updated lists or "tests" of what is spam and what is not spam. See

http://www.spamassassin.org/tests.html

for details.

Well, find out that this server has had major problems the last few days and everyone who runs SA non-locally (almost everyone) has been having these same problems. They have mirrored server to another, better location, no more problems. SA couldn't connect to that server for the "tests" list and kept trying, starting new daemons, and using all the free memory, eventually causing excessive paging to the drive.

Also discovered that I need more ram in this box.

Thanks for the suggestions.