On 30 August 2001, Philip Hazel said:
> > Can 1600 MX DNS lookups cause a 20 min
> > delay?
>
> Hmm. 0.75 seconds per lookup. That seems quite long to me. As Richard
> Welty says, it pays to have a name server on the mail host.
Yep, we use djbdns on this host -- any DNS request first goes to
dnscache on the localhost; after that it's off to the root servers.
AFAIK dnscache has no persistent state, so I suspect that if it was
recently restarted, and its cache near-empty, it could take quite a long
time to do all those MX lookups. (Actually only ~1000 of them for the
1600 list members, because that's how many unique domains are
represented in the subscriber list.)
> If you want parallelism in routing, you must give Exim multiple copies
> of the message (e.g. 16 copies with 100 recipients in each).
Ahh, maybe *that* is why Mailman by default sends to 500 recipients at a
time. (We bumped that up so all recipients would be handled in a single
SMTP session between Mailman and Exim. That might not have been such a
bright idea!)
Thanks!
Greg
--
Greg Ward - software developer gward@???
MEMS Exchange http://www.mems-exchange.org
