training Spamassassin

[pixeline]

hello,

since the upgrade to v8.0, spamassassin does not seem to catch any spam mails anymore. Besides, before the upgrades all emails i received had a header added by SA telling the score of that particular email. This is not the case anymore.

the webhost tells me SA is functioning correctly.

i found out one can train SA from within the plesk interface. Does that mean that if you don't train it it does not filter anything?

Also, is training applying only to the user's email account or to the domain mail accounts, or to the global server mail accounts ?

can someone explain the consequences of training/not training spamassassin, and what is now SA doing by default (not trained) ?

thank you, i'm sure i'm not the only one wondering what this means so your answer should be helpful to many


Alexandre

 

[misery]

I'm having trouble training spamassassin as well. I've noticed that with my main domain name I can actually login to the plesk interface and go to.. Spam Filter -> Training, from within this page, you can check box which is spam and which isn't with no problem, but if I try doing this to any other domain(s) it won't let me train it. (no blue wheel shows up).

Has anyone else encountered this problem?

 

[Thratchen]

Originally posted by misery
I'm having trouble training spamassassin as well. I've noticed that with my main domain name I can actually login to the plesk interface and go to.. Spam Filter -> Training, from within this page, you can check box which is spam and which isn't with no problem, but if I try doing this to any other domain(s) it won't let me train it. (no blue wheel shows up).

Has anyone else encountered this problem?

Actually it is possible to train spamassassin inside plesk using the spammng program


This will train for user [email protected], both the new (unread) and old (read) mails

spammng -b [email protected] -p=/var/qmail/mailnames/DOMAINNAME.COM/USERNAME/Maildir/.Spam/cur/*
spammng -b [email protected] -p=/var/qmail/mailnames/DOMAINNAME.COM/USERNAME/Maildir/.Spam/new/*

and using the same principle use this to train for ham (assuming that your inbox is free of spam)

spammng -b [email protected] -h=/var/qmail/mailnames/DOMAINNAME.COM/USERNAME/Maildir/cur/*
spammng -b [email protected] -h=/var/qmail/mailnames/DOMAINNAME.COM/USERNAME/Maildir/new/*

It should be fairly simple thing to bash script this to automatically run this every night or whatever

 

[Traged1]

The training on our servers has completely stoped working since the PLESK 8.01 upgrade.

Messages learned: 0 as spam, 0 as non-spam, 0 total.


Even right after we designate the email as spam or ham using PLESK.

ON RHEL 3 ES with PLESK 8.01

 

[Traged1]

Managed to get the Spam DB's and training back up with:

------------------------------------------------------------------------
rm -f /var/qmail/mailnames/*/*/bayes_toks
rm -f /var/qmail/mailnames/*/*/auto-whitelist
rm -f /var/qmail/mailnames/*/*/auto-whitelist.lock
rm -f /var/qmail/mailnames/*/*/bayes_seen
rm -R -f /var/qmail/mailnames/*/*/.spamassassin
rm -f /var/qmail/mailnames/*/*/user_prefs
rm -f /var/qmail/mailnames/*/*/spamcheck*

/usr/local/psa/admin/bin/mchk --with-spam
------------------------------------------------------------------------

Be aware that this will NOT UPGRADE your existing Bayes and they will be lost. However, you will get bayes functionality back and be able to train it using PLESK once more.

Also, the commands above are for RHEL only, you may have to modify the path for your OS or build.