I may not totally understand all your needs, but here is a shot from what I
glean from the post.
1) Don't host adm.uow.edu.au as a secondary on your BIND servers. You bind
servers are public, keep them that way without any knowledge of private IPs
(this is one of the problems and will continue to be a problem if using
secondaries for internal zone.) If you actually *need public IPs in the
adm.uow.edu.au zone, then create an external primary on your Bind server and
populate with required public A records (etc.) Now any external request
gets answered by your Bind server (setup a secondary on the other bind
server.) This is also called a split-horizon or split dns.
2) Have all your internal AD servers and clients point to your internal DNS
server(s) *only. If you need external rez for the clients (i.e. INET rez)
setup Forwarder to Bind server(s) for external rez. Now internal clients
will register into private space and not effect what is going on externally.
You have strong separation of duty here and boundaries. External and
internal do not mess with each other in this config.
If I am missing your problem, please set me straight and we can help you
find a solution to meet your needs. I would definitely keep w2k/w3k DNS for
your internal AD space if possible.
--
William Stacey, MS MVP
Christian Fenn said:
I work with Peter and he asked me to see if I could make the problem we're trying to solve clearer.
We have a domain adm.uow.edu.au which is an active directory domain,
delegated to the domain controller
admincat01.adm.uow.edu.au. The DC is on a private address. Our two public
addresses dns servers are acting as
secondaries for the adm.uow.edu.au zone.
The problem occurs when anyone external to our network attempts to resolve
an address in the adm.uow.edu.au zone,
eg. computera.adm.uow.edu.au. computera is on a public address. The
external client or their dns server when trying to
resolve computera.adm.uow.edu.au will, on average one time out of three, attempt to connect to
admincat01.adm.uow.edu.au which will fail (or possibly connect to a host
on their own private network with the same IP
address) Once this times out, it should then connect to one of our two
public dns servers and everything works from there.
It's the delay, due to timeout then failover, that we're trying to remove.
The way we fix the problem in our unix/bind environment is to not have an
NS record for the master for its zones, often
referred to as a "Hidden Master" arrangement. We were hoping to do the
same thing with our windows environment, but
whenever the NS record for the master is removed, it's replaced next
reload, and xfer'd back to the secondaries.