On Mon, 9 Jun 2003, David Chait wrote:
> I have been using 4.x for some time now very contently,
> however recently the university introduced a requirement that we route
> outgoing mail through one of their central smtp pools.
> On the surface that is fine, I put together a smart host to push
> mail without issue, however the pool seems to get hit pretty hard on
> occasion forcing one of their servers to refuse connections.
> Upon refusal, the messages routed to the smart host become queued,
> however do not attempt to deliver again on their own.
> Generating a lot of work on my part to clear out the queues every
> few hours. Is there any way to force Exim to redeliver the queue
> every 15 minutes or so automatically to deal with these failures?
If everything is going through a (pool of) smart host(s),
the default retry rule
* * F,2h,15m; G,16h,1h,1.5; F,4d,6h
may not be appropriate. I'm tempted to suggest trying every 15m
indefinitely
* * F,10d,15m
although I wouldn't be suprised those who know better say that you should
back off a bit.
Is the problem that the smarthosts get busy at times, or that you
are trying to push a lot through in one go ?
I'd try working with the people who run the smarthost;
they may have an idea whether it would help if you made exim
send all mails to them individually or send them in big batches down
one connection.
I'd start by tweaking "remote_max_parallel" and "queue_smtp" to try to
shape the outgoing traffic to whatever the smarthost handles best.
(Note that the remote_max_parallel docs describe a situation
where the "master" exim process should be started
exim -bd -q1m
to keep the traffic very even).
If the smarthost is particularly bad "serialize_hosts" might help,
but in that case I think the answer would be to fix the smarthost.
--
Dr. Andrew C. Aitchison Computer Officer, DPMMS, Cambridge
A.C.Aitchison@??? http://www.dpmms.cam.ac.uk/~werdna
