I'm having some trouble with Bayes.  Occasionally, mail will stop flowing
in my system.  This is because my procmail spamassassin.lock file becomes
stale.  I've futher discovered that it's because the bayes DB gets locked
and somehow dies midway during a rebuild.  This causes my whole mail system
to come to a screeching halt, and I have to go manually clear lock files
manually.

This has happened about a dozen times.  I keep trying small fixes to avoid
the problem, but it keeps popping up.  I have no idea what to do.

It just happened a few minutes ago.  Look at the timestamp on the bayes
files, and the time of the system.

[EMAIL PROTECTED] chip]$ ls -al .spamassassin/
total 14368
drwx------    2 chip     chip         4096 Nov 20 14:05 .
drwxr-xr-x   13 chip     chip         4096 Nov 20 14:12 ..
-rw-------    1 chip     chip           12 Nov 20 13:56 bayes.lock
-rw-------    1 chip     chip      1310720 Nov 20 13:56 bayes_seen
-rw-------    1 chip     chip      8392704 Nov 20 13:56 bayes_toks
-rw-------    1 chip     chip      7360512 Nov 20 13:57 bayes_toks.new
-rw-rw-r--    1 chip     chip         1213 Nov 20 00:34 user_prefs
[EMAIL PROTECTED] mail]$ date
Thu Nov 20 14:14:28 PST 2003

Spamassassin version is 2.60

[EMAIL PROTECTED] chip]$ sa-learn --dump magic
0.000          0          2          0  non-token data: bayes db version
0.000          0      12208          0  non-token data: nspam
0.000          0       1078          0  non-token data: nham
0.000          0     158284          0  non-token data: ntokens
0.000          0 1068674236          0  non-token data: oldest atime
0.000          0 1069365387          0  non-token data: newest atime
0.000          0          0          0  non-token data: last journal sync atime
0.000          0 1069366418          0  non-token data: last expiry atime
0.000          0     691200          0  non-token data: last expire atime delta
0.000          0       7514          0  non-token data: last expire reduction count

I thought the problem might be that live updates to the bayes DB was
causing lockfile contention, so I started writing everything to the bayes
journal and did an hourly rebuild.  That hasn't fixed the problem.  And
you'll notice on the timestampe about, the lockfile is NOT at 14:00 like it
should be.

I can't seem to figure out how to turn spamassasin logging on when
spamassassin is used with procmail.  How can I debug this?  Below is my
procmail configuration.  By the way, I do batch reporting with razor, and
don't use spamassassin to do the reporting.  I don't have it use a lockfile
for the razor batch, because I didn't think spamassassin -d would need any
locking, nor would it need to update bayes.

:0fw: spamassassin.lock
* < 256000
| /usr/local/bin/spamassassin 

:0 
* ^X-Spam-Status: Yes
{
        :0c
        | /usr/local/bin/spamassassin -d >> $SPAMDIR/spam.reportme-to-razor
        :0:
        $SPAMDIR/spam.to-be-archived
}

And user_prefs says:

bayes_learn_to_journal 1
bayes_journal_max_size 0


So my questions would be:

1.  Has anyone had these lockfile contention problems with bayes?
2.  How do I turn on debug output with spamassassin and procmail?
3.  How do I figure out why bayes is dying midway through an update?
4.  If I'm writing to the journal, and doing hourly rebuilds, what is
trying to rebuild the DB at 13:56?




-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive?  Does it
help you create better code?  SHARE THE LOVE, and help us help
YOU!  Click Here: http://sourceforge.net/donate/
_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk

Reply via email to