This is a discussion on Re: Spamassassin Upgrade - SpamAssassin ; Nikos K. wrote: > Dear All, > > I had a mail server with the following characteristics: > Redhat9 (Kernel 2.4.20-8) > Squirrelmail-1.4.3a-1 > Spamassassin 3.0.1. > > The mail server was working fine but the spamassassing was missing a ...
Nikos K. wrote:
> Dear All,
> I had a mail server with the following characteristics:
> Redhat9 (Kernel 2.4.20-8)
> Spamassassin 3.0.1.
> The mail server was working fine but the spamassassing was missing a lot of
> spam messages. So, I tried to upgrade it to version 3.1.8. by installing
> the following packages:
> After some more steps i ended up to:
> 1) When I run "spamassassin -V" I get: SpamAssassin version 3.1.8 running on
> Perl version 5.8.0
> 2) When I run "sa-learn --dump magic" I get:
> 0.000 0 3 0 non-token data: bayes db version
> 0.000 0 1717 0 non-token data: nspam
> 0.000 0 674 0 non-token data: nham
> 3)When I run "sa-update" I get no result, but the first time i did it, the
> following folder was created: /var/lib/spamassassin/3.001008/
That's not surprising. You'd only get "results" if there was a ruleset
update. Those don't happen every day, and sometimes months go by before
there's a rule update. Rules are created by hand, tested automatically,
and then a human decides if the rule is performing well enough and
addresses enough of a problem area to warant being published to the
updates. As a result, that's a very sporadic process, and usually a
reaction to either a very impressive rule, or a substantial mutation in
Also, SA 3.1.8 isn't a member of the current release family for SA. The
current release of SA is 3.2.3, There's a lot of development work on
getting 3.3.0 ready, and the 3.1.x branch is very minimally maintained.
The most recent updates for 3.1.x were on Sept 21, 2007 and Oct 18,
2007. The 3.2 branch got updates on Oct 17, 2007 and Oct 30 2007. So
don't expect updates every time you run sa-update. Even the current SA
doesn't get them very often.
> 4)When I run "spamassassin --lint" I get no error message.
> The PROBLEM is that spamassassin still misses a lot of spam messages and
> actually acts like before.
Ok, care to post some X-Spam-Status headers?
> Even if it seems that the new version is running now, I think that the old
> spamassassin deals with the incoming emails.
Any reason to believe that? Have you restarted spamd since your upgrade?
> Please take a look in my configuration files below:
> | /usr/bin/spamc
> required_hits 5
> rewrite_header subject ***spam***
> report_safe 0
> use_bayes 1
> bayes_auto_learn 1
> ok_locales all
> I am pretty sure that i am missing something. Is there any way to check if
> the incoming emails are handled by the new spamassassin?
Yes, look at your X-Spam-Checker-Version header.
> If indeed the new
> spamassassin handles all the incoming emails then why all this training by
> sa-update and sa-learn has given no evolution?
Dono, I'd have to at least see what rules your messages are hitting by
looking at an X-Spam-Status header. It could be you have a broken trust
path and a lot of spam is hitting ALL_TRUSTED.
You're also running what still amounts to an "old" version of
spamassassin, although I still run 3.1.x in my production environment
with some added rulesets, and I get decent results.
Are network tests enabled in your SA (ie: do you have Net:NS installed
so RCVD_FROM_* rules can be tested?)