> All BAYES_50? Silly question, but are you sure you're properly
> training?
> Running sa-learn as the right user, and all that?
> 

I must have been tired.  I thought I had run sa-learn --dump ealier, but I 
guess I didn't.  It looks like the new server has a very high ham rate and a 
low spam rate.  I'm thinking that maybe in our divine wisdom of script writing 
that someone loaded a bunch of spams using the ham script.  That's the best I 
can figure.  I checked the scripts and they are indeed using the correct user 
id.  So, I will dump the database and retrain today.  That will probably fix it.

OLD SERVER

0.000          0          3          0  non-token data: bayes db version
0.000          0    1179630          0  non-token data: nspam
0.000          0     830497          0  non-token data: nham
0.000          0     128519          0  non-token data: ntokens
0.000          0 1250654065          0  non-token data: oldest atime
0.000          0 1250780279          0  non-token data: newest atime
0.000          0          0          0  non-token data: last journal sync atime
0.000          0 1250708192          0  non-token data: last expiry atime
0.000          0      54835          0  non-token data: last expire atime delta
0.000          0      34281          0  non-token data: last expire reduction 
count

NEW SERVER

0.000          0          3          0  non-token data: bayes db version
0.000          0       5490          0  non-token data: nspam
0.000          0      10678          0  non-token data: nham
0.000          0     141755          0  non-token data: ntokens
0.000          0 1240965283          0  non-token data: oldest atime
0.000          0 1250779298          0  non-token data: newest atime
0.000          0          0          0  non-token data: last journal sync atime
0.000          0 1250735397          0  non-token data: last expiry atime
0.000          0      86400          0  non-token data: last expire atime delta
0.000          0      56262          0  non-token data: last expire reduction 
count

> All but one have subsecond scan times. Did you score an old Cray or
> something? :) That might indicate a problem, not sure.
> 
> So you have any SMTP-time DNSBL checks in place on the public MTA?
> 
> --
>   John Hardin KA7OHZ                    http://www.impsec.org/~jhardin/
>   jhar...@impsec.org    FALaholic #11174     pgpk -a jhar...@impsec.org
>   key: 0xB8732E79 -- 2D8C 34F4 6411 F507 136C  AF76 D822 E6E6 B873 2E79
> -----------------------------------------------------------------------
>    W-w-w-w-w-where did he learn to n-n-negotiate like that?
> -----------------------------------------------------------------------
>   5 days until the 1930th anniversary of the destruction of Pompeii

Reply via email to