On Mon, 21 Apr 2008, Michael Parker wrote:

> select * from bayes_vars;


....
2289 rows in set (0.00 sec)

> What user do you run bayes under on your MXs?


I think you've found the issue. We run as spamd.

# sa-learn -u spamd --dump magic
0.000 0 3 0 non-token data: bayes db version
0.000 0 1492123 0 non-token data: nspam
0.000 0 660634 0 non-token data: nham
0.000 0 73178711 0 non-token data: ntokens
0.000 0 1189775610 0 non-token data: oldest atime
0.000 0 1208785034 0 non-token data: newest atime
0.000 0 0 0 non-token data: last journal sync atime
0.000 0 0 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

That leads to two issues:

1. I need to straighten things out and figure out why I've got a
strange mix of per-user and global data in my Bayes DB. Whee.

2. Does this mean that, if I use per-user Bayes, I have to run
expiration as each user individually?

Manual expiration was recommended to me a long time ago as a way to
increase database performance, but it seems like it may not be worth
it if I have to run N forced expirations, for potentially large values
of N.

Thanks for your help.

Chris St. Pierre
Unix Systems Administrator
Nebraska Wesleyan University