On Fri, 2005-06-24 at 09:44 +0200, Ian FREISLICH wrote:
> John Horne wrote:
> > On Mon, 2005-06-20 at 12:46 +0100, Richard Hopkins wrote:
> > >
> > > What do you have spamd's "-m" (--max-children=num) set to John? On a busy
> > > mail hub you'll probably need to increase it from its default (5). We
> > > certainly did.
> > >
> > It is left to the default of 5. I thought I read in the SA docs that it
> > now couldn't be increased above 5 - I'll need to check that.
> >
> > I have changed the exim source spam.h to increase the spamd timeout from
> > 120 seconds to 240, and we have had no problems for the past hour :-)
>
> Before my patch, this timeout was 3600 seconds, and wasn't properly
> enforced. The difficulty is setting a reasonable timeout timeout
> for the connection during SMTP time. We noticed duplicate emails
> being delivered when spamd took longer than the SMTP timeout (300
> seconds) before my patch. 240 seconds is very close to 300 depending
> on what other time consuming checks you perform after DATA (maybe
> anti virus).
>
I have reverted to the original exim source with the 120 second timeout,
and increased the spamd -m option to, first 12, now 15. We actually have
two servers and once was being rebuilt so all the mail was pumped
through one machine. That obviously didn't help matters :-) I chose 240
seconds since the default smtp timeout is 5 minutes, so 240 would let
spmad timeout and still give the smtp session time to carry on. The
spamd check is the last for the data session, the mailhub will accept
the message and pass it on to another system for anti-virus checking.
> > Perhaps the spamd timeout should be a configure file option?
>
> I have thought of that, but haven't had the time yet.
>
Okay.
Many thanks,
John.
--
---------------------------------------------------------------
John Horne, University of Plymouth, UK Tel: +44 (0)1752 233914
E-mail: John.Horne@??? Fax: +44 (0)1752 233839
