[Bug 2752] spamd hang, high load, lot of memory

[bu...@bugzilla.spamassassin.org]

http://bugzilla.spamassassin.org/show_bug.cgi?id=2752





------- Additional Comments From devin.nate@bridgecomm.net  2004-01-20 20:56 -------
Our site seems to have similar problems. We use qmail-spamc->spamc->spamd. Size
limits as default, so 250k. spamd is limited to 16 processes.

local.cf: 
rewrite_subject 1
required_hits 8.0
dns_available yes
auto_whitelist_path /var/spool/spamassassin/auto-whitelist
bayes_path /var/spool/spamassassin/bayes
bayes_journal_max_size 0
bayes_expiry_max_db_size 250000
bayes_auto_expire 0
auto_learn_threshold_nonspam 1.0
auto_learn_threshold_spam 10.0
use_dcc 1
dcc_home /var/dcc
dcc_dccifd_path /var/dcc/dccifd
use_razor2 1

This machine just went through this:
  8:59pm  up 9 days, 23:15,  1 user,  load average: 1.59, 1.95, 1.95
  9:00pm  up 9 days, 23:16,  1 user,  load average: 1.91, 1.96, 1.95
  9:01pm  up 9 days, 23:17,  1 user,  load average: 1.79, 1.95, 1.95
  9:02pm  up 9 days, 23:18,  1 user,  load average: 1.53, 1.86, 1.92
  9:04pm  up 9 days, 23:21,  1 user,  load average: 31.63, 13.62, 6.27
  9:05pm  up 9 days, 23:22,  1 user,  load average: 32.89, 17.13, 7.90
  9:05pm  up 9 days, 23:22,  1 user,  load average: 32.89, 17.13, 7.90
  9:08pm  up 9 days, 23:25,  1 user,  load average: 34.20, 24.33, 12.17
  9:08pm  up 9 days, 23:25,  1 user,  load average: 34.20, 24.33, 12.17
  9:08pm  up 9 days, 23:25,  1 user,  load average: 34.20, 24.33, 12.17
  9:10pm  up 9 days, 23:27,  1 user,  load average: 26.47, 24.50, 13.74
  9:11pm  up 9 days, 23:28,  1 user,  load average: 20.85, 23.13, 13.93
  9:14pm  up 9 days, 23:31,  1 user,  load average: 25.03, 24.65, 15.94
  9:14pm  up 9 days, 23:31,  1 user,  load average: 24.63, 24.58, 15.96
  9:14pm  up 9 days, 23:31,  1 user,  load average: 23.77, 24.39, 16.00
  9:14pm  up 9 days, 23:31,  1 user,  load average: 21.22, 23.77, 15.97
  9:15pm  up 9 days, 23:31,  1 user,  load average: 18.62, 23.07, 15.87
  9:17pm  up 9 days, 23:34,  1 user,  load average: 18.35, 20.56, 15.84
  9:18pm  up 9 days, 23:35,  1 user,  load average: 20.60, 20.55, 16.17
  9:19pm  up 9 days, 23:36,  1 user,  load average: 22.72, 21.28, 16.70
  9:19pm  up 9 days, 23:36,  1 user,  load average: 21.79, 21.11, 16.67
  9:21pm  up 9 days, 23:38,  1 user,  load average: 18.79, 19.76, 16.66
  9:21pm  up 9 days, 23:38,  1 user,  load average: 18.61, 19.67, 16.68
  9:22pm  up 9 days, 23:39,  1 user,  load average: 19.11, 19.64, 16.84
  9:23pm  up 9 days, 23:40,  1 user,  load average: 15.59, 18.75, 16.65
  9:24pm  up 9 days, 23:41,  1 user,  load average: 13.69, 17.63, 16.39
  9:25pm  up 9 days, 23:41,  1 user,  load average: 7.88, 15.50, 15.73
  9:25pm  up 9 days, 23:42,  1 user,  load average: 4.75, 13.20, 14.93
  9:27pm  up 9 days, 23:43,  1 user,  load average: 2.43, 11.04, 14.08
  9:28pm  up 9 days, 23:44,  1 user,  load average: 1.06, 9.08, 13.21
  9:29pm  up 9 days, 23:45,  1 user,  load average: 0.44, 7.44, 12.38
  9:30pm  up 9 days, 23:46,  1 user,  load average: 0.52, 6.16, 11.63

According to syslog, the only things running were spamd and qmail - yes I know
syslog could be wrong - loads like these have only been observed after SA 2.6.1
was installed. sa-learn --force-expire does similar things to system load.

through cron, sa-learn --force-expire runs daily at 2am. If it's an expire run,
I don't understand why it doesn't honor the once per day limit or the config
setting saying not to do an auto-expire.

syslog output is limitted. spamd complains about not being able to get a lock on
bayes_*. dccifd complains about a broken pipe. That's about it.

Any comments?

Devin Nate




------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.