You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@spamassassin.apache.org by "help@nantucket.net" <he...@nantucket.net> on 2005/12/05 21:11:46 UTC

Bayes toks

Hey guys, the bayes_toks and bayes_seen files for my users are pretty
bloated and beginning to effect my diskspace.  Is there anything I can do ?
Just nuke them and start over? can I put a limit on them?



Re: Bayes toks

Posted by Matt Kettler <mk...@evi-inc.com>.
help@nantucket.net wrote:
> Where do I adjust this?
> 
> 
> 
> on 12/5/05 3:22 PM, Matt Kettler at mkettler@evi-inc.com wrote:
> 
> 
>>bayes_expiry_max_db_size


in your /etc/mail/spamassassin/local.cf, or any other site-level config file.

see man Mail::SpamAssassin::Conf for details on the defaults, etc.

I use a larger-than-default bayes db, and have the expiry size set to 250000.
With this I currently have about 11.4mb of bayes_toks, and my current ntokens in
sa-learn --dump magic is about 262k.

Really, with the defaults, your db should be somewhere in the 6-9mb
neighborhood, which shouldn't be "bloated" at all.


Re: Bayes toks

Posted by Pollywog <li...@shadypond.com>.
On 12/05/2005 08:26 pm, help@nantucket.net wrote:
> Where do I adjust this?
>
> on 12/5/05 3:22 PM, Matt Kettler at mkettler@evi-inc.com wrote:
> > bayes_expiry_max_db_size

http://spamassassin.apache.org/full/2.6x/dist/doc/sa-learn.html

   has the details

Re: Bayes toks

Posted by "help@nantucket.net" <he...@nantucket.net>.
Where do I adjust this?



on 12/5/05 3:22 PM, Matt Kettler at mkettler@evi-inc.com wrote:

> bayes_expiry_max_db_size



Re: Bayes toks

Posted by Matt Kettler <mk...@evi-inc.com>.
help@nantucket.net wrote:
> Hey guys, the bayes_toks and bayes_seen files for my users are pretty
> bloated and beginning to effect my diskspace.  Is there anything I can do ?
> Just nuke them and start over? can I put a limit on them?

For bayes_toks, you can do sa-learn --force-expire, and the size limit is
defined by bayes_expiry_max_db_size, which adjusts the number of tokens SA keeps.

For bayes_seen, you'll have to nuke it. Note that this isn't entirely safe in
3.0.x and earlier, but in 3.1.x they added code to allow it to be nuked.

See http://issues.apache.org/SpamAssassin/show_bug.cgi?id=2975