You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@spamassassin.apache.org by Paco Yepes <pa...@um.es> on 2006/01/05 14:38:57 UTC

no bayes check applied in MTA

Hello.

Im using postfix 2.1.5 + spamassassin 3.1.0a-2 + amavisd-new
20030616p10-5

In the last weeks the size of false-negatives (fn) is grow
significatelly. All this fn do not have scores of type BAYES_XX (and I
think that is not normal).

When I check the same message (fn) with spamc, then the message is
detected like spam.

Anybody know what is happening?

Thanks.


-- 
Francisco Yepes Candel
Sección de Telemática
ATICA
Universidad de Murcia
pacoy@um.es
Tlf: +34 968 364828
Fax: +34 968 398337


Re: no bayes check applied in MTA

Posted by Paco Yepes <pa...@um.es>.
El vie, 06-01-2006 a las 09:55 +0100, mouss escribió:
> Paco Yepes a écrit :
> > Hello.
> > 
> > Im using postfix 2.1.5 + spamassassin 3.1.0a-2 + amavisd-new
> > 20030616p10-5
> > 
> > In the last weeks the size of false-negatives (fn) is grow
> > significatelly. All this fn do not have scores of type BAYES_XX (and I
> > think that is not normal).
> > 
> > When I check the same message (fn) with spamc, then the message is
> > detected like spam.
> > 
> > Anybody know what is happening?
> 
> Try running spamc as the same user as amavisd-new (vscan, clamav, amavis
> or whatever it is on your system) and see the results.
> 
> if you run sa-learn, do the same thing.

I not detect any error when executing spamc as root or amavis ... the
same false negative is detected like spam in the two cases.

I was received too a response of Jim Gottlieb:

> I was having the same problem, and the log showed that the Bayes
> database was often locked from another process so it couldn't be
> accessed.
> 
> I switched to having the Bayes database in MySQL and now it's much
> better.

But there is not logs that show any type of lock in the bayes bd ...

Could be a problem with the number of spam messages learned (>20000)?:

# sa-learn --dump magic
0.000          0          3          0  non-token data: bayes db version
0.000          0      21985          0  non-token data: nspam
0.000          0       2219          0  non-token data: nham
0.000          0     113841          0  non-token data: ntokens
0.000          0 1136266108          0  non-token data: oldest atime
0.000          0 1137150086          0  non-token data: newest atime
0.000          0 1137150092          0  non-token data: last journal
sync atime
0.000          0 1137129908          0  non-token data: last expiry
atime
0.000          0     691200          0  non-token data: last expire
atime delta
0.000          0       4367          0  non-token data: last expire
reduction count

An other stuff: the only parameter of amavisd.conf I was changed is $sa_local_tests_only. It is now = 0 (false).
Since spamassassin is running with this setting the number of false negatives has decreased.

Some idea about what's happening?

Thanks.

-- 
Francisco Yepes Candel
Sección de Telemática
ATICA
Universidad de Murcia
pacoy@um.es
Tlf: +34 968 364828
Fax: +34 968 398337