SA timeout help.

[Hancock, Scott]

I've been looking for a SA timeout issue for several days.  Is the
numberign supposed to start at 0 and not 1?  From the mail log


Aug 10 10:39:38 pebbles MailScanner[31668]: SpamAssassin timed out and
was killed, failure 0 of 20
Aug 10 10:39:39 pebbles MailScanner[31668]: Message 1GBBdG-0002OT-Sq
from 209.200.5.12
(3-5968161-morganco.com?narainv at intqw.turnedtheold22.com) to
morganco.com is not spam, SpamAssassin (timed out)

I did find several problems but now when I run check_mailscanner and
scan the output, I notice only two issues.

The first is check_mailscanner can take a over a minute to get past the
line 

[24665] dbg: bayes: expiry max exponent: 9

in the following.


[24665] dbg: bayes: expiry starting
[24665] dbg: locker: refresh_lock: refresh
/var/lib/MailScanner/bayes.lock
[24665] dbg: locker: refresh_lock: refresh
/var/lib/MailScanner/bayes.lock
[24665] dbg: bayes: DB expiry: tokens in DB: 16432117, Expiry max size:
150000, Oldest atime: 1116109310, Newest atime: 1155039924, Last expire:
1116253906, Current time: 1155044021
[24665] dbg: bayes: expiry check keep size, 0.75 * max: 112500
[24665] dbg: bayes: token count: 16432117, final goal reduction size:
16319617
[24665] dbg: bayes: first pass? current: 1155044021, Last: 1116253906,
atime: 144604, count: 36214, newdelta: 320, ratio: 450.643867012757,
period: 43200
[24665] dbg: bayes: can't use estimation method for expiry, unexpected
result, calculating optimal atime delta (first pass)
[24665] dbg: bayes: expiry max exponent: 9
[25485] dbg: message: ---- MIME PARSER START ----
[25485] dbg: message: main message type: multipart/mixed
[25485] dbg: message: parsing multipart, got boundary:
----=_NextPart_001_0000_01C6BAFD.3D6884F0
[25485] dbg: message: found part of type text/plain, boundary:
----=_NextPart_001_0000_01C6BAFD.3D6884F0
[25485] dbg: message: parsing normal part
[25485] dbg: message: added part, type: text/plain
[25485] dbg: message: ---- MIME PARSER END ----



The second problem is a permissions problem in the Bayes folder.  The
journal file is owned by root and the rest of the files are owned by
mail.  The mail process does not run as root.  At other points in the
check_mailscanner output, Bayes entries are entered successfully.
check_mailscanner output does not pause at all when the permissions
error appears.

See line 

[29627] dbg: locker: refresh_lock: refresh
/var/lib/MailScanner/bayes.lock
bayes: bad permissions on journal, can't read:
/var/lib/MailScanner/bayes_journal

below

[29627] dbg: bayes: opportunistic call found expiry due
[29627] dbg: bayes: bayes journal sync starting
[29627] dbg: locker: refresh_lock: refresh
/var/lib/MailScanner/bayes.lock
[29627] dbg: locker: refresh_lock: refresh
/var/lib/MailScanner/bayes.lock
bayes: bad permissions on journal, can't read:
/var/lib/MailScanner/bayes_journal
[29627] dbg: bayes: bayes journal sync completed
[29627] dbg: bayes: expiry starting
[29627] dbg: locker: refresh_lock: refresh
/var/lib/MailScanner/bayes.lock
[29627] dbg: locker: refresh_lock: refresh
/var/lib/MailScanner/bayes.lock
[29627] dbg: bayes: DB expiry: tokens in DB: 16432241, Expiry max size:
150000, Oldest atime: 1116109310, Newest atime: 1155211501, Last expire:
1116253906, Current time: 1155218450
[29627] dbg: bayes: expiry check keep size, 0.75 * max: 112500
[29627] dbg: bayes: token count: 16432241, final goal reduction size:
16319741
[29627] dbg: bayes: first pass? current: 1155218450, Last: 1116253906,
atime: 144604, count: 36214, newdelta: 320, ratio: 450.647291102888,
period: 43200
[29627] dbg: bayes: can't use estimation method for expiry, unexpected
result, calculating optimal atime delta (first pass)
[29627] dbg: bayes: expiry max exponent: 9


Could either of these issues result in SA timeouts?  
Any fix suggestions?  
Should this post be on the SA list?

I have the same problem on two mailscanners.  One is running the latest
MS version from the tar package in /opt.  The other running the latest
Debian package.  

FWIW, I made an honest attempt at understanding the debain packaging
system to make my own Debian package.  Running from /opt was much more
simple.

Thanks for any help or pointers.

Scott