MailScanner lock up Problem

[Paul Kelly :: Blacknight Solutions]

Hi All,

One of our boxes has developed a weird lock up problem in the last 24 hours.

Version: 4.63.8
SA: 3.2.3
Perl: 5.8.5

Once started MailScanner will process a few 100 mails and then simply
lock up.

e.g.

mail     22296  0.2  1.3 63840 58056 ?       S    14:03   0:06
MailScanner: dangerous content scanning
mail     22325  0.3  1.4 64288 58524 ?       S    14:03   0:06
MailScanner: waiting for messages
mail     22384  0.2  1.3 63104 57344 ?       S    14:04   0:05
MailScanner: checking with SpamAssassin
mail     22452  0.3  1.4 64412 58632 ?       S    14:04   0:06
MailScanner: waiting for messages
mail     22506  0.3  1.4 64364 58520 ?       S    14:04   0:06
MailScanner: finishing batch
mail     22543  0.2  1.3 62660 56900 ?       S    14:04   0:05
MailScanner: checking with SpamAssassin
mail     22588  0.2  1.3 63504 57652 ?       S    14:04   0:05
MailScanner: spam checks
mail     22608  0.2  1.4 64212 58372 ?       S    14:04   0:06
MailScanner: waiting for messages
mail     21668  0.0  0.5 26948 21112 ?       Ss   14:03   0:00
MailScanner: master waiting for children, sleeping
mail     21669  0.2  1.3 63088 57272 ?       S    14:03   0:05
MailScanner: waiting for messages
mail     22176  0.2  1.4 64648 58836 ?       S    14:03   0:06
MailScanner: dangerous content scanning


They will stay like that for ever. If you do a restart, the processes
never die and the init script is unhappy and keeps saying:

Waiting for MailScanner to die gracefully
...................................


it'll keep going like that, till I have to force a restart.

An strace on some of the PIDs shows:

Parent process:

# strace -p 21668
Process 21668 attached - interrupt to quit
waitpid(-1,  <unfinished ...>

Children:

# strace -p 21669
Process 21669 attached - interrupt to quit
write(5, "<22>Sep  7 14:08:37 MailScanner["..., 77 <unfinished ...>

# strace -p 22176
Process 22176 attached - interrupt to quit
write(5, "<22>Sep  7 14:19:50 MailScanner["..., 113 <unfinished ...>

No other activity from them. We're running this on CentOS 4.5 and most
of the important perl modules are current.

If i run MailScanner --debug --debug-sa, the process exists once it's
batch is done with no obvious errors or anything.

Any of you got any ideas??

thanks,

Paul

-- 
Paul Kelly
Technical Director
Blacknight Internet Solutions ltd
Hosting, Colocation, Dedicated servers
IP Transit Services
Lo-call: 1850 929 929
DDI: 059 9183091

e-mail: paul at blacknight.ie
web: http://www.blacknight.ie

Blacknight Internet Solutions Ltd,
Unit 12A,Barrowside Business Park,
Sleaty Road,
Graiguecullen,
Carlow,
Ireland

Company No.: 370845