hits=0.0

[jberliner <sa...@spamtrench.com>]

I'm running a patched qmailrocks-type setup: qmail + qmail-scanner 1.25 +
ClamAV. 

I'm having a problem where at times about half of our messages are getting
0.0 scores from SA, even ones that I know should be triggering rules. I even
wrote a really stupid custom rule (test for the presence of the letter 'e'
in message body) and it doesn't always fire.

Message headers of emails that get a 0.0 score always look like:
X-Spam-Status: No, hits=0.0 required=4.1
X-Spam-Report: SA TESTS
 _SUMMARY_

instead of, say:
X-Spam-Status: No, hits=-3.5 required=4.1
X-Spam-Report: SA TESTS
 -1.8 ALL_TRUSTED         Passed through trusted hosts only via SMTP
  0.1 JMB_MATCH_ALL     BODY: Test for the letter e!
 -2.6 BAYES_00             BODY: Bayesian spam probability is 0 to 1%
                                  [score: 0.0000]
  0.8 AWL                    AWL: From: address is in the auto white-list

There doesn't seem to be any rhyme or reason or pattern that I can see to
which emails hit rules and which don't, which suggests maybe something based
on CPU or network load, e.g. spamd timing out? However we are not under very
high load -- typical CPU load is around 0.01 and rarely exceeds 0.1. We
process roughly 2-3000 messages a day for a single domain, about half of
which are probably spam.

I am loading a lot of plugins though: Bayes, URIDNSBL, Hashcash, DCC, Pyzor,
Razor2, AWL, VBounce, etc.

spamd invocation =
# spamd -x -u qscand -H /tmp -s /dev/stderr -m10

(I just upped max-children today to 10 as above, but without much apparent
effect.)

Any clues to help troubleshoot? Random similar posts elsewhere have
suggested problems with AWL or corrupted Bayes db, but timing out sounds
most likely to this novice sysadmin...
-- 
View this message in context: http://www.nabble.com/hits%3D0.0-tp21382998p21382998.html
Sent from the SpamAssassin - Users mailing list archive at Nabble.com.