postfix: Process did not exit cleanly, returned 1 with signal 0

[Dave Jenkins]

2008/6/6 Glenn Steen <glenn.steen at gmail.com>:
> 2008/6/5 Dave Jenkins <davejenx at googlemail.com>:
>> MailScanner & postfix have mostly been running happily for a few
>> weeks, averaging about 5 msgs/min. But on two occasions we've had
>> defunct MailScanner processes and the error "postfix: Process did not
>> exit cleanly, returned 1 with signal 0".
...
>> ...Finally I turned off
>> scanning (Scan Messages = no) and this allowed the queue to clear.
>>
>> I then switched scanning, virus and spamassassin back on (i.e.
>> restored previous config) and it then ran fine. I'm wondering if it
>> was a peculiarity of one of the messages that caused MailScanner to
>> crash.
>>
>> The second time it happened, before clearing the queue I took a copy
>> of /var/spool/MailScanner/incoming; will that help in debugging the
>> problem? If so, what should I do with it and if not, what should I do
>> next time it happens?
> ...
> Next time it happens, look in the hold queue, the oldest <batch
> size... usually 30 or so> messages, to see if you have a problem
> there.

So it's the Incoming Queue Dir, in my case /var/spool/postfix/hold,
that I should grab a copy of, rather than Incoming Work Dir
(/var/spool/MailScanner/incoming), is that right?

>  If it were non-queue files fouling things up, doing what you
> did would likely not have cleared things up, so it might not be the
> usual razor agent log misplaced in the hold directory...

That makes sense to me; I would have thought the fact that disabling
scanning alowed the queue to be processed successfully and re-enabling
scanning restored normal function, suggests that MailScanner choked
during the scanning of a specific message. So I'm guessing the defunct
MailScanner processes were the cause rather than effect of the postfix
process exiting uncleanly.

All files in my copy of /var/spool/MailScanner/incoming were owned by
postfix:postfix and all files perm 600, subdirectories 700.

> More like some subtle permission thing or subcomponent "borking totally" on a
> specific mail.
> Trick is to find what part and what to do with it:-):-)

I think I've found which mail caused the problem. By finding the time
of the first "postfix: Process did not exit cleanly" in
/var/log/messages then looking at what happened around that time in
/var/log/maillog, I've found a message that fits the timing perfectly
and that seems to be the first in a growing queue of messages that got
repeatedly but unsuccessfully processed until I set Scan Messages to
off. I recognise the recipient from my perusing of the logs the first
time the problem happened. But this time I have the offending
message's entry from /var/spool/MailScanner/incoming:

-rw------- 1 postfix postfix 1014 Jun  5 15:39 9E379FE019.AC06E.header

9E379FE019.AC06E:
total 32
drwx------  2 postfix postfix  4096 Jun  5 15:39 .
drwx------ 31 postfix postfix  4096 Jun  5 15:39 ..
-rw-------  1 postfix postfix  5571 Jun  5 15:39 msg-1101-41.txt
-rw-------  1 postfix postfix 14362 Jun  5 15:39 msg-1101-42.html

When I view the files with less, I get:
"9E379FE019.AC06E/msg-1101-41.txt" may be a binary file.  See it anyway?

# file 9E379FE019.AC06E/msg-1101-41.txt
9E379FE019.AC06E/msg-1101-41.txt: Microsoft Office Document

The .html file is OK.

Cheers,

Dave