I'm not quite sure how to describe this problem; if I could, I'd
probably have it solved.  So I'll go into the symptoms.

For about a year spamassassin has worked pretty much flawlessly, with
the occasional hiccup attributable to my own mistakes.

I run spamassassin from qmail-scanner as email gets stuffed into the
qmail queue; spamassassin 2.43, daemonized.  Here's my local.cf:

rewrite_subject         0
skip_rbl_checks         1
use_terse_report        1
report_header           1
spam_level_char         x
defang_mime             0

I'm running Debian, mostly woody, but with some stuff from sid,
including libc6.  I haven't done any upgrades in the last week or so,
during which spamassassin has run quite happily.  I upgraded to
linux-2.4.20 last week sometime as well, but with no major config
changes - I added USB support, but the module wasn't loaded.
perl version 5.6.1 (debian revision 8.2).

This morning I woke up with a system load around thirty; my tcpserver (I
run qmail) smtp process having spawned the maximum number of child
processes, and (because of the high load) dozens of gotmail processes
running, and duplicate emails being sent because of it.

looking at my process listing, something similar to this was seen, but
multiplied by thirty or so:

qmaild     446  0.0  0.0  1268  484 ?        S    10:16   0:00 /usr/bin/tcpserver -R 
-H -u 64011 -g 65534 -v -x /etc/tcp.smtp.cdb 0 smtp /usr/sbin/relay-ctrl-check 
/usr/sbin/qmail-smtpd
root      3752  0.0  0.0  1356  460 ?        S    10:26   0:00  \_ 
/usr/sbin/qmail-smtpd
qmailq    4065  0.0  0.6  4900 3616 ?        S    10:26   0:00  |   \_ /usr/bin/perl 
/usr/local/bin/qmail-scanner-queue.pl
root      4634  0.0  0.0  1288  300 ?        S    10:29   0:00  \_ 
/usr/sbin/qmail-smtpd
root      4986  0.2  0.0  1356  460 ?        S    10:30   0:00  \_ 
/usr/sbin/qmail-smtpd
qmailq    4990  5.0  0.7  4976 3804 ?        S    10:30   0:00      \_ /usr/bin/perl 
/usr/local/bin/qmail-scanner-queue.pl
qmailq    4994  0.0  0.1  2108 1020 ?        S    10:30   0:00          \_ sh -c 
/usr/bin/spamc  -f < /var/spool/qmailscan/working/new/neko10421370514244990
qmailq    4995  0.3  0.1  1860  532 ?        S    10:30   0:00             \_ 
/usr/bin/spamc -f

none of the spamc processes seemed to be letting go of control.
Curious.  spamd isn't forking subprocesses either.

There were no error messages of note in the log files.

I killed spamd, and waited a few minutes for the dust to clear, to at
least allow email to go through.

Restarting spamassassin made mail start to back up again, with the same
problem.  spamc wouldn't complete; spamd isn't spawning any subdaemons
to handle the requests, it appears.

I telnetted to the spamd port.  It's listening.  Typed in some random
stuff.  No errors, but I don't know the spamd protocol.

On a hunch I backed up and deleted the autowhitelist file, since it had
(!!) ballooned up to six megabytes.

Also turned off the autowhitelist option in /etc/default/spamassassin:

OPTIONS="-i 127.0.0.1"

Out of ideas, I rebooted my machine.  Still no dice.

Any suggestions?

I have it turned off for the interim, and I've got a few unhappy users
because of it. :)

-- 
Rob Bos, rbos@{novylen.net,constainia.net}, upstanding atheist

"Goodbye, goodbye,
        I don't care 
 If I never taste your fine food again,
 neutral fellows, seers of every side.
 Tolerance, what crimes 
 are committed in your name."
                -- Denise Levertov, _Goodbye To Tolerance_

Attachment: msg11912/pgp00000.pgp
Description: PGP signature

Reply via email to