Re: Troubleshooting Bayes

On Wed, 2008-03-12 at 11:11 -0700, Kurt Buff wrote:
>    X-Spam-Status: No, hits=0 tagged_above=-999
required=5
>     tests=DKIM_SIGNED=0.001, DKIM_VERIFIED=-0.001,
DK_SIGNED=0.001,
>     SPF_PASS=-0.001
>    X-Spam-Level: 
> 
> It's definitely possible I'm looking at something
incorrectly, but that's what I found so far.

I think bayes needs a minimum amount of ham and spam to
start tagging.
If you're still not seeing bayes tags, check the db table
directly,
bayes_vars, and make sure you counts are rising. This is
also important
to watch as you don't want spam/ham counts lopsided.

-- 
Robert

_______________________________________________
Maia-users mailing list
Maia-usersrenaissoft.com
http://www.renaissoft.com/mailman/listinfo/maia-users

Robert Fitzpatrick wrote:
> On Wed, 2008-03-12 at 11:11 -0700, Kurt Buff wrote:
>   
>>    X-Spam-Status: No, hits=0 tagged_above=-999
required=5
>>     tests=DKIM_SIGNED=0.001, DKIM_VERIFIED=-0.001,
DK_SIGNED=0.001,
>>     SPF_PASS=-0.001
>>    X-Spam-Level: 
>>
>> It's definitely possible I'm looking at something
incorrectly, but that's what I found so far.
>>     
>
> I think bayes needs a minimum amount of ham and spam to
start tagging.
> If you're still not seeing bayes tags, check the db
table directly,
> bayes_vars, and make sure you counts are rising. This
is also important
> to watch as you don't want spam/ham counts lopsided.
>   
Bayes needs 200 ham and 200 spam before it will be used for
scoring. I 
understand more than 2000 before its really useful.

You might try (as amavisd user) sa-learn --dump magic. Check
that nham 
and nspam are over 200. Even then, every email isn't going
to fire a 
bayes score. I think I would get some mail into at least ham
cache, run 
process-quarantine and check the output so you know
something made it 
into bayes, then --dump magic. It is critical that all parts
of the 
system are interacting with bayes as the same user
(override_user 
notwithstanding). My method is the "can't write a
proper select count 
query" method 

0.000          0          3          0  non-token data:
bayes db version
0.000          0     244301          0  non-token data:
nspam
0.000          0      51666          0  non-token data:
nham
0.000          0     390214          0  non-token data:
ntokens
0.000          0 1204029253          0  non-token data:
oldest atime
0.000          0 1205464646          0  non-token data:
newest atime
0.000          0          0          0  non-token data: last
journal 
sync atime
0.000          0 1205126102          0  non-token data: last
expiry atime
0.000          0     345600          0  non-token data: last
expire 
atime delta
0.000          0       1079          0  non-token data: last
expire 
reduction count

_______________________________________________
Maia-users mailing list
Maia-usersrenaissoft.com
http://www.renaissoft.com/mailman/listinfo/maia-users