SpamAssassin is not reporting any Bayes score

fredgc · 05-03-2009, 07:12 PM

I have a gentoo computer I am using as the household firewall, mail server, etc. It is running Amavis-new with SpamAssissin.

SpamAssissin is filtering spam. The built in rules are being reported and my local rules are also working.

The problem is that no messages get a Bayes score, neither positive nor negative. Can anybody recommend ways to debug this?

The log file shows it is accessing the database:

Code:

jeeves ~ # grep -i bayes /var/log/mail | tail
May  3 12:17:06 jeeves amavis[22342]: (22342-18) SA dbg: locker: safe_lock: created /home/spamtrap/.spamassassin/bayes\.lock.jeeves.gchouse.org.22342
May  3 12:17:06 jeeves amavis[22342]: (22342-18) SA dbg: locker:safe_lock: trying to get lock on /home/spamtrap/.spam\assassin/bayes with 0 retries
May  3 12:17:06 jeeves amavis[22342]: (22342-18) SA dbg: locker:safe_lock: link to /home/spamtrap/.spamassassin/bayes\.lock: link ok
May  3 12:17:06 jeeves amavis[22342]: (22342-18) SA dbg: bayes: tie-ing to DB file R/W /home/spamtrap/.spamassassin/bayes_toks
May  3 12:17:07 jeeves amavis[22342]: (22342-18) SA dbg: bayes: tie-ing to DB file R/W /home/spamtrap/.spamassassin/bayes_seen
May  3 12:17:08 jeeves amavis[22342]: (22342-18) SA dbg: bayes: found bayes db version 3
May  3 12:17:13 jeeves amavis[22342]: (22342-18) SA dbg: bayes: learned '28345cc8de2fbaacc7baee1447b61e27ed2c3514@sa_generated', atime: 1241378209
May  3 12:17:13 jeeves amavis[22342]: (22342-18) SA dbg: bayes: untie-ing
May  3 12:17:14 jeeves amavis[22342]: (22342-18) SA dbg: bayes: files locked, now unlocking lock
May  3 12:17:14 jeeves amavis[22342]: (22342-18) SA dbg: locker: safe_unlock: unlink /home/spamtrap/.spamassassin/bayes.lock

The database does look like its growing:

Code:

jeeves ~ # sudo -H -u spamtrap sa-learn --dump magic
0.000          0          3          0  non-token data: bayes db version
0.000          0        792          0  non-token data: nspam
0.000          0       2255          0  non-token data: nham
0.000          0     137433          0  non-token data: ntokens
0.000          0 1234689516          0  non-token data: oldest atime
0.000          0 1241378209          0  non-token data: newest atime
0.000          0          0          0  non-token data: last journal sync atime
0.000          0 1240218428          0  non-token data: last expiry atime
0.000          0    5529600          0  non-token data: last expire atime delta
0.000          0      51402          0  non-token data: last expire reduction count

And finally, amavisd is running as user spamtrap:

Code:

jeeves ~ # ps auxw | grep amavis
spamtrap   533  0.1 18.6  62848 23508 ?        S    12:56   0:14 amavisd (ch8-avail)
spamtrap   952  0.1 20.1  62276 25420 ?        S    13:28   0:13 amavisd (ch6-avail)
root      3520  0.0  0.4   1760   596 pts/0    S+   16:48   0:00 grep --colour=auto amavis
spamtrap 22327  0.0  1.6  60948  2052 ?        Ss   02:06   0:14 amavisd (master)

What did I forget?

rweaver · 05-04-2009, 02:38 PM

Try adding this to your local.cf

use_bayes 1
bayes_auto_learn 1
add_header all Status _YESNO_, score=_SCORE_ required=_REQD_ tests=_TESTS_ autolearn=_AUTOLEARN_ version=_VERSION_

See if that looks a bit more interesting.

fredgc · 05-05-2009, 12:14 AM

I already had
use_bayes 1
bayes_auto_learn 1
So I am adding the third line. Now I guess I wait for some mail and look at the logs...

archtoad6 · 06-08-2009, 01:04 PM

What happened? -- Any improvement?

fredgc · 06-10-2009, 12:05 AM

Thank you for asking!

I'm still not seeing anything in /var/log/mail except what I had above about connecting to the database. I suspect that I shouldn't?

However, I now see some messages with BAYES_00=-2.599, and a few with BAYES_20. I think that means it is working at finding ham (it was correct -- the email was indeed ham). But I'm not seeing anything in my junk folder marked with BAYES. Could my database be bad?

There are some messages in the junk folder. These are all messages that matched home-grown rules for forged headers.

Any advice?

billymayday · 06-10-2009, 12:34 AM

A couple of points

First, if you are using amavisd, add_headers will have no effect, since amavisd adds headers, not spamassassin.

Second, I suspect that header checks are taking place before content filtering, and hence not being sent to amavisd.

Try sending yourself a test spam message that won't get caught by your header checks and look at the headers. See http://spamassassin.apache.org/gtube/