> Rocco, >> I have ported naive bayes spam DB -> MySQL. I can't see any improvement at >> the moment... Should the CPU works less then when I use naive db (I noted >> the bayes_seen file in home dir of amavis user was about 100 MB and similar >> dim has the auto-whitelist file, so I think that thi can be an high load >> for the CPU to examine every time that one message needs to be scanned - >> also my system send/receive about 110.000 message per day)?
You can (and on occasion should) trim down the AWL file. If you imported the AWL data into SQL, then you can: log onto MySQL USE <your bayes database>; DELETE FROM awl WHERE count="1"; If it is not in MySQL, you can use the check_whitelist program supplied in the tools directory of the source code, or wget: http://www.deepnet.cx/~kdeugau/spamtools/trim_whitelist Run these programs as your amavisd-new user. With your bayes_seen file growing that large, I wonder if you are syncing and expiring your Bayes data on a regular basis (this should happen automatically, you would have had to disable this function for it not to). You can see if your are by checking the number of tokens before and after a --sync --force-expire. To check before (assuming your amavisd-new user is 'vscan'): sa-learn -u vscan --dump magic Then clean the database (if possible): su vscan -c 'sa-learn --sync --force-expire' Then check again: sa-learn -u vscan --dump magic If the before and after numbers vary by a large amount, then you need to insure you run the --sync --force-expire (as the amavisd-new user) on a daily basis. Gary V ------------------------------------------------------- This SF.Net email is sponsored by: Power Architecture Resource Center: Free content, downloads, discussions, and more. http://solutions.newsforge.com/ibmarch.tmpl _______________________________________________ AMaViS-user mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/amavis-user AMaViS-FAQ:http://www.amavis.org/amavis-faq.php3 AMaViS-HowTos:http://www.amavis.org/howto/
