On Montag, 15. Mai 2006 17:01 Theo Van Dinter wrote: > Since 209328+1347841 < 2000000, an expire wouldn't have been > attempted opportunistically. However, running "sa-learn > --force-expire" would have made it happen.
Yes, this was the cause. I have bayes_auto_expire 0 therefore a nightly --force-expire job. > 209328+1347841 = 1557169. > Bayes will try to expire tokens down to 75% of the max number. So > 2000000*0.75 = 1500000, which means it saw that it should try to > expire 1557169-1500000 = 57169 tokens. That would have been expected. > If the previous expire looks similar to this expire run, Bayes will > use an estimate for atime to do the expire. This speeds things up, > but also could possibly be wrong (it is an estimate afterall). So if > an estimate was used, up to 1557169-100000 tokens (100k is a safety > net) could be expired. I still don't understand the reason for that. Wouldn't it be bad to throw away that many tokens? mfg zmi -- // Michael Monnerie, Ing.BSc ----- http://it-management.at // Tel: 0660/4156531 .network.your.ideas. // PGP Key: "lynx -source http://zmi.at/zmi3.asc | gpg --import" // Fingerprint: 44A3 C1EC B71E C71A B4C2 9AA6 C818 847C 55CB A4EE // Keyserver: www.keyserver.net Key-ID: 0x55CBA4EE
pgpDgXHkQG5BF.pgp
Description: PGP signature
