How to config to do this about bayes?

[Monty Ree <ch...@hotmail.com>]

Hello, all.

I have setup bayes at local.cf below and works well.0

use_bayes               1
bayes_path /var/spool/spam/.spamassassin/bayes
bayes_file_mode 777
bayes_auto_learn        1

but after that, bayes* files increasing continuously.
I'm afraid that this would make SA runs more slowly.

So I would like this.
Is this possible? 
If then(just one of the three or together?) which config is required? 


-. bayes files are limited just 10M bytes with the lastest information. 

-. or number of spam or ham are limited just 100,000 with the lastest 
information.

-. or bayes db information is limited within one month.

 

Thanks for your time...

_________________________________________________________________
보다 빠른 소식, 보다 빠른 정보, MSN 뉴스에서 확인하세요. 
http://news.msn.co.kr/ 

Re: How to config to do this about bayes?

[Matt Kettler <mk...@verizon.net>]

Monty Ree wrote:
> Hello, all.
>
> I have setup bayes at local.cf below and works well.0
>
> use_bayes 1
> bayes_path /var/spool/spam/.spamassassin/bayes
> bayes_file_mode 777
> bayes_auto_learn 1
>
> but after that, bayes* files increasing continuously.
> I'm afraid that this would make SA runs more slowly.
>
> So I would like this.
> Is this possible? If then(just one of the three or together?) which
> config is required?
>
> -. bayes files are limited just 10M bytes with the lastest information.
> -. or number of spam or ham are limited just 100,000 with the lastest
> information.
>
> -. or bayes db information is limited within one month.

SA has none of the above options.

However, you can limit it by token count, and this is the way it is by
default. SA automatically checks to see if there's lots of tokens in the
database and tries to do an expiry run to reduce their numbers by
removing some of the older tokens. By default, SA tries to keep between
100,000 and 150,000 tokens in the database. You can increase that with
the bayes_expiry_max_db_size option, but SA will never expire to less
than 100,000 tokens. Such a small database wouldn't be very useful. This
should result in a bayes DB somewhere around 12-16MB (the docs say 8MB,
but that's based on an outdated format)

If your bayes DB is getting huge, are there a bunch of files ending in
".expire" laying about?

If so:
- run sa-learn --force-expire to manually cause an expiry run. This may
take several minutes to finish
- Find out what part of your toolchain is killing SA while it's trying
to do bayes maintenance. If you use MailScanner, it's default config
commonly causes this.