[Mimedefang] defunct defang processes

[David F. Skoll]

On Sun, 6 Mar 2005, John Nemeth wrote:

> } 	I have installed sendmail/mimedefang/spamassasin on a solaris system.
> } It seems to be working OK, but some of the defang processes created by
> } the mimedefang program stay around even the mail for which the process
> } was created has been processed and it's subdirectory in
> } /var/spool/MIMEDefang has been removed. These processes are marked as
> } defunct. 

>      Hmm, this is interesting.  I'm seeing the same thing on NetBSD
> 2.0.  Since this is the first release of NetBSD with native pthreads, I
> thought it was an OS issue, but maybe not.  I ended up writing a script
> that uses md-mx-ctrl slaves to find out which slaves MIMEDefang knows
> about and kill the rest, which is run periodically from cron.

That doesn't make too much sense.  Are you seeing defunct "mimedefang"
threads, or defunct "mimedefang.pl" Perl processes?  If the former, then
doing anything with md-mx-ctrl is useless, because the mimedefang threads
are controlled by libmilter, and the multiplexor knows nothing about them.

If you're seeing defunct mimedefang.pl slaves, then there's a serious
bug somewhere -- the multiplexor is supposed to catch SIGCHLD and reap
terminated slaves.  I strongly doubt this is failing, because if it
were failing, the multiplexor would eventually fail completely because
it would be unwilling to start new slaves.  If a slave hasn't been
reaped yet, the multiplexor thinks it's still running.

Regards,

David.