Bayes expire timeouts?

[Peter Bonivart]

I stopped the automatic token expire by adding "bayes_auto_expire 0" to
spam.assassin.prefs.conf. I instead run the following command from
crontab every night.

 > Your "cron" job on kleenex
 > /usr/local/bin/sa-learn --force-expire --rebuild
 >
 > produced the following output:
 >
 > ..
 >
 >
...........................................................................................................................................................................................................................................................synced
 >  Bayes databases from journal in 1 seconds: 2049 unique entries (3428
 > total
 > entries)
 > expired old Bayes database entries in 35 seconds
 > 165430 entries kept, 86110 deleted
 > token frequency: 1-occurence tokens: 54.00%
 > token frequency: less than 8 occurrences: 32.25%

As you can see it took 35 seconds to do the job. The SA timeout is 40
seconds so it's not unlikely that some of the SA timeouts are caused by
SA trying to expire tokens. Now I run this command during low traffic
hours to try to minimize the problem.

I have had the Bayes database get corrupt (or something) a couple of
times. Trying to access it causes the R/W error and rebuild doesn't
help, I have to do an import which fixes the database but I have to wait
for 200 spams before it starts scoring again.

Does anyone have more info about this? I think the Bayes scores are most
important and want it to work in a more reliable way.

--
/Peter Bonivart

--Unix lovers do it in the Sun

Sun Fire V210, Solaris 9, Sendmail 8.12.10, MailScanner 4.25-14,
SpamAssassin 2.61 + DCC 1.2.21, ClamAV 0.65 + GMP