Hello,
Two questions really:
1) At present the Uni has one central mailhub with a disaster recovery plan
should that system fall over. I'm not too happy with that (too much work on
my part! :-) ), and am trying to configure Exim to work with a 'split mail
hub' (not sure if that's a correct term or not :-) ) Basically, I want it to
work a bit like the DNS with a master/slave system. I think this is possible
since nearly all our data (users, etc) is held in cdb files. These can be
copied to the slave system when they are rebuilt. No problem. The question I
guess is has anyone else done this (rather than just a load-balancing between
systems), and if so are there any gotchas to be aware of? The only one I can
see is at our site is that the master system will have to handle the
majordomo mailing lists since users may modify the lists without the slave
knowing anything about it (and I'm not going to hack majordomo to do this).
However, the slave system will have a final 'unknown user' director (which
will also catch any majordomo requests) and send them to the 'master' system.
2) We used to get ask quite a bit, although not so much now for some reason,
the name and/or IP address of the mailhub so that users could configure
their MUAs/MTAs when sending receiving mail. Typically this was "It's
asking for our local SMTP server - what's that?". We provided a DNS entry
(called 'mail') which they could use. It was an A record with the same IP
address of the mailhub and a single MX record pointing back to itself.
However, with the split mailhub I'd like this to point to either system. My
first thought was to remove the A record and just use MX records, but that
won't work since the MUA will no doubt be sending mail just to the IP address
rather than looking up MX records (which an MTA would(?) do). Instead could
I do something like:
mail IN A 141.163.2.9
IN A 141.163.2.62 (added for the split mail hub)
IN MX 8 mail
Then - I think - anything wanting an MX record will see that it points to
itself and just one of the A records. Likewise if it wants an actual IP
address, then it can use either of them. Does this look okay? Sorry the
question is a bit off-topic. (I guess multi-homed machines do something like
this? but in our case its two separate machines.)
Thanks,
John.
------------------------------------------------------------------------
John Horne, University of Plymouth, UK Tel: +44 (0)1752 233914
E-mail: jhorne@???
PGP key available from public key servers
