Title: Site link costs
Not actually knowing anything about your environment, I can
only make some general comments, but if your goal of controlling "how often and
with who each site replicates" is your reason for disabling the KCC then that's
not really a good reason. The KCC works off the topology you define.
If it's creating connection objects between sites you don't want then it's
because a) someone misconfigured the topology or b) there was an outage and the
KCC deemed it necessary to work around the outage to maintain the flow of
replication. The latter option is dependent upon if 'bridge all site
links' is enabled, if you have redundant site links, etc. Now, there are
some instances when manual connection objects might be called for, but it should
definitely be the exception, not the norm. Not saying you're one of these,
but I normally ask customers who do this type of thing if they're bored or have
a lot of time on their hands. Since they've normally already told me they
lack the time and resources to be doing all the things they should be doing,
they normally can't provide a satisfactory answer. :) But again, I
don't know your environment or situation, so take my comments as generic and not
necessarily applicable to you.
Great
question, we just had this at our place. We just finished deploying a W2K3
AD structure across the globule with each division using their own sub
domain.
We are
creating our site links manually. And by saying "We" I mean one of the
five Enterprise admins across the globe. We have deactivated the ability
for AD to create its own links so we don't have to worry about
oddities.
The
reason for this is so we can control how often and WITH WHO each site
replicates. Right now we have the site that hosts the first DC for each
domain replicating back to sites with root domain controllers but all other
domain sites only replicate with each other and their first DC. This means
that if the link between our root domain controllers and that primary domain
controller site was to go away we wouldn't have replication with
them.
The
links that were being created by AD weren't what we wanted. We had sites
in Italy replicating with New Jersey and sites in Mexico replication with
Ireland. I think this had something to do with our routing tables,
firewall placements and frame relay clouds that we are using across the
globe.
So, I
guess it all depends on your topology that you have.
Charlie
Sorry for the basic question...
Our company just upgraded our NT4 domains in-place
as child W2K3 domains under an empty W2K3 forest root domain. 22 sites and
their associated subnets were established, with one subsidiary leaving all
their objects in the default first site because they feel their bandwidth will
support it. However, we're currently having heated discussions regarding AD
and site topology.
Some IT members are saying that there is no need to
manually create site links or assign properties such as cost and replication
interval. They say that if we don't do this, then AD does it automatically and
it will do a better job than we would anyway.
I thought that the KCC needed the site
topology info to be provided (whether manually or programmatically) so that it
could automatically create the connection objects (provided you're not
manually creating them).
So who is confused here, me or them? This should be
basic stuff, and I want to understand it correctly :-).
TIA,
Cathy