Mailscanner allocating swap memory space and not freeing it?

[Kai Schaetzl]

During the last few days I have seen several occasions where SA timed out 
(and I'm sure it is RBLs), so that MS couldn't add a score and delivered 
the messages as clean. It's always accompanied by those "attempt 1 of 20" 
messages (never higher than that).
After investigating I found that if that occurs our swap space is heavily 
in use although there's still plenty of RAM available. "heavily in use" 
means that top and free show 200 or more MB of swap space in use while 
that usually is near nothing. As an example:

             total       used       free     shared    buffers     cached
Mem:        514948     364916     150032          0      13400      51128
-/+ buffers/cache:     300388     214560
Swap:       514040     261756     252284

Also, an "sa-learn -D --sync" takes a long time, around two minutes (which 
explains the time-out). I didn't check the size of the journal file the 
first time I forced the sync, but this time I did and it was at 170 MB (!) 
(almost ten times higher than the bayes db itself). I'd think that's 
impossible on a machine with only a few thousand mails arriving per day. 
Looks like a problem with the journal in SA 3.0. That also explains why it 
cuts into the swap space. What it doesn't explain is why that space isn't 
freed up after the time-out. It gets only freed when I restart 
MailScanner.
Before I post this to sa-talk I wanted to check here what MailScanner 
actually does when SA notices a --sync is needed (I assume it does one 
each day?) and why the swap space isn't redeemed after that.


Kai

-- 

Kai Schätzl, Berlin, Germany
Get your web at Conactive Internet Services: http://www.conactive.com
IE-Center: http://ie5.de & http://msie.winware.org

------------------------ MailScanner list ------------------------
To unsubscribe, email jiscmail at jiscmail.ac.uk with the words:
'leave mailscanner' in the body of the email.
Before posting, read the MAQ (http://www.mailscanner.biz/maq/) and
the archives (http://www.jiscmail.ac.uk/lists/mailscanner.html).