On Wed, 29 Aug 2012 09:40:29 +0200 Rob Sterenborg wrote: > Hello, > > I'm in the process of installing a new relay MTA, and part of it > means moving the current Bayes database to the new server (bdb -> > MySQL). > > The database is quite big because of the bayes_token and bayes_seen > tables, so I wanted to expire old data. I've set > bayes_expiry_max_db_size=1000000 in local.cf[1] (there were *way* > more entries) and ran "sa-learn -D --force-expire -u > '<username>'"
-u is for virtual users with per user bayes, you should run this as the unix user that runs spamd or equivalent. > What bothers me is that I can't update the > spam_count and ham_count fields because AFAIK I don't have > information about that. You shoudn't normally touch those in an expire. > Because bayes_seen was also quite big I read up on that too. Since > the table doesn't include any age information and (most) everything I > found says "just delete it", I emptied the table. Although I think > it's strange to just throw away information about previous seen > messages that have been classified as either spam or ham. Any other > insight in this would be valued.. Some people include atime information for that purpose.
