Well, a few more information : Output of sa-learn --dump magic -D :
[22420] dbg: bayes: found bayes db version 3 [22420] dbg: bayes: DB journal sync: last sync: 1157102359 [22420] dbg: config: score set 3 chosen. 0.000 0 3 0 non-token data: bayes db version 0.000 0 1189366 0 non-token data: nspam 0.000 0 197582 0 non-token data: nham 0.000 0 387408 0 non-token data: ntokens 0.000 0 1157049872 0 non-token data: oldest atime 0.000 0 1157102360 0 non-token data: newest atime 0.000 0 1157102359 0 non-token data: last journal sync atime 0.000 0 1157093142 0 non-token data: last expiry atime 0.000 0 43200 0 non-token data: last expire atime delta 0.000 0 295143 0 non-token data: last expire reduction count [22420] dbg: bayes: untie-ing [22420] dbg: bayes: untie-ing db_toks [22420] dbg: bayes: untie-ing db_seen If I read well, there's 387408 tokens in the DB... Despite there's no bayes_expiry_max_db_size specified anywhere and the defalut value is 150000 (??) Shall I issue a sa-learn --force-expire command ? Does it supposed to work ? Stephane -----Message d'origine----- De : Stéphane LEPREVOST [mailto:[EMAIL PROTECTED] Envoyé : vendredi 1 septembre 2006 10:18 À : 'users@spamassassin.apache.org' Objet : RE: Very big auto-whitelist file One more question in the same way : my bayes_seen file is quite huge too (about 160Mb) Googling around about this I saw there was some bugs with versions prior to 3.1 but despite I'm using version 3.1.1 (a bit late on upgrading too, I'm afraid :-\ ) I think there's something wrong here too... Is there a way to fix it or to trim the file ? Stephane -----Message d'origine----- De : Stéphane LEPREVOST [mailto:[EMAIL PROTECTED] Envoyé : jeudi 31 août 2006 22:39 À : 'users@spamassassin.apache.org' Objet : RE: Very big auto-whitelist file Thanks Kris for this usefull tool, I'll try it tommorow (and thanks to Roger too who noticed the existence of your tool) As you noticed, I get worried very very very late... But in fact I wasn't in charge of spamassassin when we first saw this growth, that's why I'm back on the problem only now... I guess I'll pay more attention to this now ;D Stephane -----Message d'origine----- De : Kris Deugau [mailto:[EMAIL PROTECTED] Envoyé : jeudi 31 août 2006 21:58 À : users@spamassassin.apache.org Objet : Re: Very big auto-whitelist file Roger Taranto wrote: > There's an additional tool to run after you run check_whitelist. It's > called trim_whitelist, and it compacts the db file. I can't remember > where I found it, but you should be able to google for it. It should > reduce the size of your db file quite a bit. That would be the ancient creaky tool I wrote ~2 years ago. <g> Make sure to read the notes and caveats regarding DB_File/AnyDBM_File. Google seems to have lost, or *very* heavily downrated, the direct link to the space I posted it (and a few other tools) to, so: http://www.deepnet.cx/~kdeugau/spamtools/ And I wrote it because of this exact problem of AWL files growing indefinitely... although I got worried around 5M instead of 1.2G. ;) -kgd
