Re: Little help...

Sean Conner <[email protected]> Sat, 5 Apr 2008 19:39:33 -0400
Newsgroups gmane.mail.spam.greylist.user
Message-ID <[email protected]>
It was thus said that the Great Paul Venezia once stated:
> On Apr 4, 2008, at 11:35 PM, RayJ wrote:
> >
> > I am having an intermittent problem with the relaydelay daemon (
> > relaydelay.pl ). It tends to crash a few time a day at random. As a
> > quick fix I created a cron to monitor the PID and, if needed, restart
> > relaydelay.pl. This "stop gap" solution is working, but I have been
> > unable to pin down why the crash in the first place.
> >
> I've noticed this as well, but only on CentOS/RHEL. The same code was  =

> running fine on a FC3 box prior to an upgrade. It may have something  =

> to do with the sendmail-milter, but I haven't had time to look into it.

  I've had problems with milter for my own greylist implementation [1] which
I suspect has to do with stack space.  Milter creates a bunch of threads,
and at least under Cent-OS (which we use at work), uses about 10M per stack
per thread.  Currently, I explicitely set the stack limit (using the system
call setrlimit(); also doable from the command line via ulimit) to a smaller
value and that seems to have helped.

  -spc (How small?  Try 128k stack limit)

[1]	http://www.x-grey.com/