[Mimedefang] SA/MIMEDefang performance tweaking

[Chris Gauch]

Did the same thing from my end -- enabled "learn_to_journal" and set my
bayes DB to auto-expire when it grows to about 50M.

bayes_expiry_max_db_size 50000000

It also helps to adjust the auto-learning thresholds (so that your bayes DB
won't grow NEARLY as fast).  I don't bother learning email as spam unless it
scores above 25 points.  For ham, email must score 0.5 or lower.  

bayes_auto_learn_threshold_nonspam 0.50
bayes_auto_learn_threshold_spam 25.00

It's worked out well for almost a week now, no .lock file issues at all and
Bayes is working out quite well.  The only drawback to this method of
auto-learning is that Bayes will only be 90-100% confident for obvious spam
(since it's learning spam from email scoring so highly).

- Chris 



-----Original Message-----
From: mimedefang-bounces at lists.roaringpenguin.com
[mailto:mimedefang-bounces at lists.roaringpenguin.com] On Behalf Of Lucas
Albers
Sent: Saturday, June 12, 2004 2:58 PM
To: mimedefang at lists.roaringpenguin.com
Subject: RE: [Mimedefang] SA/MIMEDefang performance tweaking


Matthew.van.Eerde at hbinc.com said:
>> From: Chris Gauch [mailto:cgauch at digicon.net]
>> Is there any way to continue to use Bayes WITHOUT autolearn
>
I had the lock file problem.
Just configure learn_to_journal and run a background sa-learn --rebuild
job every hour.
That will solve the lock problems.
I encountered your exact problem on my external, and that solved it!


-- 
Luke Computer Science System Administrator
Security Administrator,College of Engineering
Montana State University-Bozeman,Montana

_______________________________________________
Visit http://www.mimedefang.org and http://www.canit.ca
MIMEDefang mailing list
MIMEDefang at lists.roaringpenguin.com
http://lists.roaringpenguin.com/mailman/listinfo/mimedefang