Scenario:
* spamc (2.60) on a power-pc aix 4.1.3.0 is invoked ("spamc -t 300 -c -u spamd
-d <server> -p <port>") by qmail to query:
* spamd (from spamassassin 2.60) on a Linux 2.4.21-0.13mdksmp 2 X Intel(R)
Xeon(TM) CPU 2.40GHz, started by daemontools,
* Bot machines on the same 100 MHz Ethernet segment.
Spamd is running nicely, taking about 12 seconds to service each spamc
-request. There are about 100 mails pr. minute. This spins along for maybe 3
hours.
At 18:54 (09/30, CEST, that's zulu+2h) *something* happens, so that most of
the spamd-processes, started around that point in time, takes incredibly
long. You can see from the log that some of them finish more than two hours
after 18:54:
------------------------------------------------
21:12:39 identified spam (12.9/8.0) for spamd:500 in 8692.3 seconds, 5151
bytes.
21:13:11 clean message (2.0/8.0) for spamd:500 in 7861.8 seconds, 157185
bytes.
21:13:49 clean message (6.8/8.0) for spamd:500 in 8409.3 seconds, 3683 bytes.
21:14:01 clean message (5.0/8.0) for spamd:500 in 8718.5 seconds, 7843 bytes.
21:15:09 clean message (0.8/8.0) for spamd:500 in 8405.0 seconds, 991 bytes.
-------------------------------------------------
(No pattern: Bytesize or spam/ham, seems visible to me).
All of the waiting spamd-processes appears to be in the run-queue, thus
driving the load-factor way above 200. (At which point the command-line
becomes extremely sluggish, the oxygen-masks drop from above and I reboot
:-). The spamc-process on the client has usually timed out, due to the "-t
300" option.
I don't know what happend at 18:54, but i dont think it's the load, since the
amount of mail usually drops off before 17:00.
Around the time, there are log-entries like this "Cannot open bayes databases
/home/spamd/.spamassassin/bayes_* R/W: lock failed: File exists" - but I
don't know if this is connected. I did remember to run "sa-learn --import"
when upgrading to 2.60.
(I've been running the spamd-server with debug ("-D") option for a few hours
now, but of course without reproducing the problem.)
Any hints? How can spamd take so long when there are plenty of default
timeouts in the the setup?
TiA, Ole Nomann Thomsen
-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
