bayes DB not growing

[Kristian Shaw]

    [ The following text is in the "iso-8859-1" character set. ]
    [ Your display is set for the "US-ASCII" character set.  ]
    [ Some characters may be displayed incorrectly. ]

Hello,

I experienced a similar problem this week - my bayes database didn't appear
to be learning new messages even though I wasn't getting any errors. The
database was well established (>6 months old) and I was seeing bayes scores
in spam reports.

In the end I did a backup of the bayes database, removed the bayes files
from ~/.spamassassin and then did a restore.

After that, I've been seeing learning activity (nham/nspam increasing). I
assume that the backup/restore rebuilt a clean database without whatever
error it had before.

Regards,

Kris.

----- Original Message -----
From: "Arif Malik" <Arifm at TOMASJEWELRY.COM>
To: <MAILSCANNER at JISCMAIL.AC.UK>
Sent: Wednesday, May 18, 2005 6:44 PM
Subject: Re: bayes DB not growing


thanks for the reply, Martin! I checked all the confs I could find, and
the only one I found is identical to what you pasted below... still I
get every message showing "autolearn=spam". Shouldn't anything under 12
points be marked as "autolearn=no"? and anything under .1
"autolearn=ham" ?? when I run sa-learn --dump magic, it looks like the
number of spams is not growing, so it doesn't appear to actually be
feeding it to bayes...

 I just did a test also to see if anything is being fed to bayes, and
changed the bayes_auto_learn_threshold_spam to 4 - since every message I
am receiving currently is spam anyways, I figured I could have that
number low - but when I sa-learn --dump magic the number of spams is not
growing. And still, every message is tagged with "autolearn=spam".

-----Original Message-----
From: MailScanner mailing list [mailto:MAILSCANNER at JISCMAIL.AC.UK] On
Behalf Of Martin Hepworth
Sent: Wednesday, May 18, 2005 1:24 AM
To: MAILSCANNER at JISCMAIL.AC.UK
Subject: Re: bayes DB not growing

Looks like the autolearn threshhold is too low to me.

the defaults in /usr/local/share/spamassassin/10_misc.cf are

bayes_auto_learn_threshold_nonspam   0.1
bayes_auto_learn_threshold_spam              12.0

I'd make sure you're not overriding these values in any of your site
specific rules in /etc/mail/spamassassin or spam.assassin.prefs.conf

--
Martin Hepworth
Snr Systems Administrator
Solid State Logic
Tel: +44 (0)1865 842300


Arif Malik wrote:
>  ok sorry, now that I am using the bayes starter kit thing, I do have
> something like this in my logs:
> BAYES_95 2.06
>
> so that part looks good to me - but now I am just worried about the
> autolearn=spam thing - from what I have read, messages tagged like
> this are fed to the bayes filter as spam... but EVERY one of my
> messages is showing this, even the ones that arent tagged as spam...
> so I am worried the ham messages are being fed to it as well.
>

------------------------ MailScanner list ------------------------
To unsubscribe, email jiscmail at jiscmail.ac.uk with the words:
'leave mailscanner' in the body of the email.
Before posting, read the Wiki (http://wiki.mailscanner.info/) and
the archives (http://www.jiscmail.ac.uk/lists/mailscanner.html).

Support MailScanner development - buy the book off the website!

------------------------ MailScanner list ------------------------
To unsubscribe, email jiscmail at jiscmail.ac.uk with the words:
'leave mailscanner' in the body of the email.
Before posting, read the Wiki (http://wiki.mailscanner.info/) and
the archives (http://www.jiscmail.ac.uk/lists/mailscanner.html).

Support MailScanner development - buy the book off the website!