Weblog entry #5 for oxtan
See all weblog entries made by oxtan.
#5
spamassassin troubles
Posted by oxtan on Mon 6 Mar 2006 at 10:52
A few weeks ago I installed and configured spamassassin with postfix/amavis-new. It all worked well, until I added too much ham to the bayes database.
According to the sa-learn man page, you can have messages forgotten with the option --forget , but I must be doing something really wrong because it does not work for me.
Let's see:
# sa-learn --dbpath /var/lib/amavis/.spamassassin/bayes --dump magic
0.000 0 3 0 non-token data: bayes db version
0.000 0 242 0 non-token data: nspam
0.000 0 1910 0 non-token data: nham
0.000 0 69428 0 non-token data: ntokens
0.000 0 1108734233 0 non-token data: oldest atime
0.000 0 1141641613 0 non-token data: newest atime
0.000 0 0 0 non-token data: last journal sync atime
0.000 0 1141536305 0 non-token data: last expiry atime
0.000 0 0 0 non-token data: last expire atime delta
0.000 0 0 0 non-token data: last expire reduction count
as you can see, the number of ham is way higher than nspam, I would like to reverse that. In order to achieve it I then do:
# sa-learn --forget /path/to/maildir/cur/*
runs fine but it will not forget the ham, if I run the --dump magic command I get similar results.
Is there a way in the debian package to start from scratch? I tried with dpkg-reconfigure spamassassin but that did not help either. I honestly do not know how to proceed from here. Any help greatly appreciated
According to the sa-learn man page, you can have messages forgotten with the option --forget , but I must be doing something really wrong because it does not work for me.
Let's see:
# sa-learn --dbpath /var/lib/amavis/.spamassassin/bayes --dump magic
0.000 0 3 0 non-token data: bayes db version
0.000 0 242 0 non-token data: nspam
0.000 0 1910 0 non-token data: nham
0.000 0 69428 0 non-token data: ntokens
0.000 0 1108734233 0 non-token data: oldest atime
0.000 0 1141641613 0 non-token data: newest atime
0.000 0 0 0 non-token data: last journal sync atime
0.000 0 1141536305 0 non-token data: last expiry atime
0.000 0 0 0 non-token data: last expire atime delta
0.000 0 0 0 non-token data: last expire reduction count
as you can see, the number of ham is way higher than nspam, I would like to reverse that. In order to achieve it I then do:
# sa-learn --forget /path/to/maildir/cur/*
runs fine but it will not forget the ham, if I run the --dump magic command I get similar results.
Is there a way in the debian package to start from scratch? I tried with dpkg-reconfigure spamassassin but that did not help either. I honestly do not know how to proceed from here. Any help greatly appreciated
Comments on this Entry
Posted by Steve (212.20.xx.xx) on Mon 6 Mar 2006 at 14:44
[ Send Message | View Steve's Scratchpad | View Weblogs ]
[ Send Message | View Steve's Scratchpad | View Weblogs ]
I'd guess that the results of the training are stored in a file somewhere, and that removing that database/file would reset your training back to the default.
Unfortunately I don't know where that might be.
You could look in /var, your home directory, or use a debugging tool such as strace to see where things are loaded from ..
[ Parent | Reply to this comment ]
thanks for the strace pointer, I should have thought of that myself, actually. Just the day before I used it for another tool. I'll see what I find.
[ Parent | Reply to this comment ]