Weird behaviour after disabling sa-learn

[Scott Ryan <sc...@staff.telkomsa.net>]

Hi, I have been doing some testing with SA - Using maildrop to do the spam 
scanning. 
In my maildrop script I was playing around with calling 'sa-learn --sync spam' 
everytime spam was detected and 'sa-learn --sync ham' when messages were 
clean. I had this running for a while to see what kind of impact/improvement 
this had if any. I then came to the conclusion that there seemed to be no 
point in teaching bayes spam when SA allready knows its spam, so I disabled 
the sa-learn calls. All good.
Until I looked at my spam graphs - (generated with qmailmrtg7). I have 
attached my graph so you can see the precise moment that I disabled sa-learn. 
The load on the machine plummetted (expected) - but the graphs also indicated 
that the number of mails scanned (clean + spam) had significantly decreased. 
I am hoping that this is some type of error with qmailmrtg7 reporting, 
because obviously (to management) something looks seriously wrong.

Any advice here would be appreciated 
-- 
Regards,

Scott Ryan
Telkom Internet
-------------------------------------
Good judgement comes with experience. 
Unfortunately, the experience
usually comes from bad judgement.
-------------------------------------