Establishing Site Links question when you have a mesh network?

[cajuntank]

Just got done with an upgrade to our WAN about a month ago to MetroE circuits for my school district. I have multiple sites that currently have 100Mb circuits each and my district office having a 500Mb circuit to aggregate to.

I have currently all sites under the defaultipsitelink default of 100 cost and 15 minutes replication frequency. KCC in it's "wizdom" has chosen one of my oldest servers at one of my 100Mb sites as it's main server. I am about to decommision that box altogether with a new 2008 DC.

Is there a better way that I need to be defining this to address the lower cost of my district office site? Also, when I do "decom" the old server, is there anything I need to watch/wait for in regards to KCC picking another box to make the "main go to" box before I add another DC?

As always, TIA.

 

[cajuntank]

Bueller?... Bueller?

 

[kmcferrin]

KCC in it's "wizdom" has chosen one of my oldest servers at one of my 100Mb sites as it's main server.

I'm a little confused by what you mean by "it's main server." Are you saying that KCC has named one of your oldest servers as the bridgehead for that site? If so you could manually set a preference order for the bridgehead role and just not include the old, soon to be decommissioned server as a bridgehead.

There isn't really a "main server" as I understand it. Your central site will probably be the administrative center of your network, and so you will probably hold all of the FSMO roles there. But each site should have it's own DCs and GCs in case a network link goes down.

As far as AD replication goes, the traditional model would be a hub and spoke model, where each of the satellite sites replicates only with the main site. That may be what you mean by "main server"?

At any rate, an AD Site is essentially defined as a set of well-connected links. When you're dealing with 100Mbps and faster WAN links, you're talking WANs that are as fast as many of us were using for LAN links originally. You could technically probably get away with having just one big site instead of multiple sites, and letting replication and authentication go willy-nilly. At least until you lose a site link.

So for purposes of availability I would stick with the old hub and spoke model. However, there is nothing wrong with specifying multiple remote bridgeheads for each satellite, but leaving the "central hub" as the preferred partner. This causes replication to follow the hub and spoke model, unless the server at the hub is unavailable, in which case the spokes can still replicate with other spokes, preventing isolation.

KCC is going to run by default every 15 minutes and pick the most efficient replication design available. I don't know how the weighting works within a site, but it may be that the oldest DC in that site (or the one with the longest uptime?) gets priority. If you take that server offline then KCC will find another replication bridgehead for that site. Or, as I said, you can set them yourself and not use KCC and ISTG.

By the way, what is the functional level of your forest? The ISTG algorithm has been improved several times as newer versions of Windows Serer have been released, but they all use the version specified by the forest functional level. Newer versions will of course be more effective.

________________________________________
CompTIA A+, Network+, Server+, Security+
MCTS:Hyper-V
MCTS:System Center Virtual Machine Manager
MCSE:Security 2003
MCITP:Enterprise Administrator

 

[cajuntank]

I'm saying that when I drill down to each of my servers, the replication partner for all of my servers at all of my sites is this one lowly server that I'm about to retire. I would prefer KCC "automatically" choose my FSMO server since it's at my higher bandwidth site.

The functional level for my forest is 2003. I will soon, in about another month or two, have all of my DC(s) at 2008 at which time, I'll raise functional level to 2008.

 

[kmcferrin]

I see. I would first double-check that all of your sites re configured correctly and the correct servers are in each site.

Then I would verify that all FSMO roles are held by servers other than the decomm target.

Then I would manually configure the preferred bridgehead to be used for site replication so that all spoke sites replicate with the hub site.

Then I would decomm that old server.

________________________________________
CompTIA A+, Network+, Server+, Security+
MCTS:Hyper-V
MCTS:System Center Virtual Machine Manager
MCSE:Security 2003
MCITP:Enterprise Administrator