On 2/9/2013 1:49 PM, Wietse Venema wrote:
> Whatever command pipe(8) executes, it will have to
>
> 1) terminate
> AND
> 2) close stdout and stderr.
>
> If that command does not do both then pipe(8) will wait (up to a
> configurable time limit after which it kills the process).
spamassassin unix - n n - - pipe
user=nobody argv=/usr/bin/spamc -f -e
/usr/sbin/sendmail -oi -f ${sender} ${recipient}
Ok, so spamc is being called to pipe the msg to spamd, and should
terminate after doing so. When finished processing the msg, spamd pipes
the msg back into Postfix via the sendmail compatibility command, correct?
I just tracked the processes via top with a test msg. pipe spawned,
spamc spawned and terminated instantly, I saw the message in my MUA
within 5 seconds, and somewhere around a minute later the pipe process
terminated.
So where do I go from here? Is there some way I can instrument
something to gather some relevant data when this delay occurs next?
Having ~1% of mail delayed 3-4 minutes isn't life threatening. I
wouldn't have even known there was a problem if not for sending a msg to
a list and waiting to confirm it went out. I'd sure like to fix it
anyway. I'm guessing the problem is with spamd. But I don't know if
it's due to bad configuration on my part, a problem with spamd itself, a
Debian issue, etc. I don't find any warnings or errors anywhere in the
logs.
--
Stan
