On Thu, 24 Oct 1996, Neal Becker wrote:
> Why would I have loads of exim processes all individually trying to
> connect to the same remote? I thought exim was designed to avoid
> this?
Not specifically. If n individual messages to the same host arrive on
your system, one by one, then n individual processes will be started to
deliver them, unless you have set one of the queue_only options. If each
one takes a long time to deliver or time out, then you might well see n
processes connected to the remote machine.
Exim is primarily designed for use on systems directly connected to the
Internet. In our environment we find that 98% of messages can be
delivered first time, with very little delay, most of the time. That is
why Exim is the way it is, with minimum administrative overhead.
It is only when a message fails to deliver to a remote host that a
record is made that the host is down, in which case subsequent messages
get queued up for the next retry and, if possible, sent down the same
SMTP connection.
In an environment when many connections time out, or are otherwise
delayed, Exim's default strategy is not the best. In that situation it
might be better to set the queue_smtp option so that remote deliveries
are initially put on the queue. These will subsequently be picked up by
a queue runner process, and all those for the same host will go down the
same SMTP connection.
--
Philip Hazel University Computing Service,
ph10@??? New Museums Site, Cambridge CB2 3QG,
P.Hazel@??? England. Phone: +44 1223 334714
