spamassassin tuning

[jdow]

From: "Justin Zygmont" <jzygmont at solarflow.net>

> does anyone know if there is a solution for a problem with spam messages 
> getting through like this, it doesn't even register a score.
> 
> 
> X-Spam-Checker-Version: SpamAssassin 3.1.1 (2006-03-10) on
>  solarflow.dyndns.org
> X-Spam-Level: 
> X-Spam-Status: No, score=0.0 required=1.8 tests=HTML_MESSAGE autolearn=failed
>  version=3.1.1
> 
> 
> X-Spam-Checker-Version: SpamAssassin 3.1.1 (2006-03-10) on
>  solarflow.dyndns.org
> X-Spam-Level: 
> X-Spam-Status: No, score=0.0 required=1.8 tests=HTML_MESSAGE autolearn=failed
>  version=3.1.1

It looks like the SpamAssassin setup is (typically) Fedoraed and
does not work right as you have it configured.

At least it can find its rules. But that looks like about all it
can find. So let's start at the beginning with how you are running
it. Is it running as a service? If so what options do you have
in /etc/sysconfig/spamassassin? If there is no file there check
SPAMDOPTIONS in the /etc/init.d/spamassassin file. If it is running
through Amavis-D I can't help but maybe somebody else can. Part of
the idea is to figure out what user SpamAssassin is run under.
(Important note, SpamAssassin WILL NOT run as root. It instantly
drops to nobody if you try. ALWAYS do mail as a user rather than
as root.) I suppose we also need to know if you are trying to use
SQL or not and whether you are trying for global Bayes or 
individual Bayes. (If you can afford the resources use the latter.
It will ultimately work better.)

Then we need to look at what you have in /etc/mail/spamassassin/*.
That can tell is something of where it thinks it is living.

Finally you would need to take a typical message and feed it to
SpamAssassin with the -D option. SpamAssassin is simply a filter.
So you can run it "SpamAssassin -D < message" to see what it is
doing. You can also use the -t option to see a score recap at the
end. When you do this run SpamAssassin as the user it is running
as when it is testing mail. That will tell you a whole lot about
your setup, what can and cannot be found.

(If you use spamd and spamc you can feed messages to spamc the same
way except that the -t option doesn't work.)

I am guessing that the user as which SpamAssassin is running cannot
access the default bayes directory. So you have NO Bayes scoring
at all.

You also have your scoring set to 1.8. This is severely fscked. Put it
back to stock, 5. Train Bayes. THEN worry about the threshold score.
Once you get Bayes working we can discuss the SARE rule sets and their
use. But let's get it working the way you intend, first.

{^_^}