Periodic (5min) SpamAssassin timeouts

[Tony Johansson]

We're having problems with periodic SpamAssassin timeouts.
We have 3 Centos 4.2 servers, running MailScanner 4.54.6 and SpamAssassin 3.1.5

The timeouts seem to hit servers individually, with numerous timeouts all 
coming in 5 minute (roughly) intervalls.

No 5 minute cronjobs exists, timeouts seem to occur and stop for no apparent 
reason. Loads, network connectivity etc are the same for all 3 servers, yet 
timeouts aren't evenly distributed among the servers when they occur - always 
1 or rarely 2 servers get hit.

dcc and pyzor are disabled in spam.assassin.prefs.conf
SpamAssassin Timeout = 180 in MailScanner.conf

I've set ut a swatch-job that execs spamassassin --lint -D, iostat, top -b -n 
1, vmstat, netstat -p whenever the timeouts occur. So far I havent been able 
to spot something that sticks out.

Heres a cut from tonights log:
Oct 10 03:12:29 mx3 MailScanner[31635]: SpamAssassin timed out and was killed 
(pid 6174), failure 0 of 20
Oct 10 03:17:30 mx3 MailScanner[31959]: SpamAssassin timed out and was killed 
(pid 8844), failure 0 of 20
Oct 10 03:22:33 mx3 MailScanner[32069]: SpamAssassin timed out and was killed 
(pid 11596), failure 0 of 20
Oct 10 03:30:11 mx3 MailScanner[32069]: SpamAssassin timed out and was killed 
(pid 15825), failure 0 of 20
Oct 10 03:35:15 mx3 MailScanner[32421]: SpamAssassin timed out and was killed 
(pid 18582), failure 0 of 20
Oct 10 03:40:16 mx3 MailScanner[31679]: SpamAssassin timed out and was killed 
(pid 21353), failure 0 of 20
Oct 10 03:45:20 mx3 MailScanner[32171]: SpamAssassin timed out and was killed 
(pid 24035), failure 0 of 20
Oct 10 03:50:23 mx3 MailScanner[31959]: SpamAssassin timed out and was killed 
(pid 26559), failure 0 of 20
Oct 10 03:55:23 mx3 MailScanner[31885]: SpamAssassin timed out and was killed 
(pid 28997), failure 0 of 20
Oct 10 04:00:29 mx3 MailScanner[31215]: SpamAssassin timed out and was killed 
(pid 31610), failure 1 of 20
Oct 10 04:05:31 mx3 MailScanner[30786]: SpamAssassin timed out and was killed 
(pid 1662), failure 1 of 20

No timeouts after 04:05. (pid info added to aid debugging)

Any ideas on whats going on here? 

Any tips on how to further debug this?

Regards, Tony