My “unofficial” Exchange Server 2007 CCR recommendations
Having had Cluster Continuous Replication in place now for about a year, I’ve derived some of my own best practice recommendations that might not be found within the official Microsoft documentation. Of course, it’s been a little while since I last checked that documentation, so this could be duplication of recommendations.
Although these recommendations are based on a very solid and stable production environment…anything bad or job search inducing that happens as a result of implementing them in your environment is not my responsibility.
So, here is the list in order of importance:
- Hub Transport servers should NOT be the location for your File Share Witness. I have found the BEST location in my opinion is in fact Domain Controllers in the same domain as the Exchange servers. This is obviously to be extended to cover other sites and domains as your organization may have them. Of course, you need to set up two FSW locations…one primary and one staged with folders and permissions in place in the event your primary FSW server dies.
- I do know for a fact that the original guidance Microsoft gave was to have the FSW on the Hub Transport servers. The logic was that a Hub Transport is a required role and the FSW needs to be on a server the Exchange administrator controls…a natural fit.
- The servers hosting the File Share Witness (or at least the one server that is the primary FSW host) should NOT be in the same data center as either of the two Mailbox server nodes in the CCR cluster. By using a three data center model (all local and in the same AD site) you provide complete and automatic fail over protection in the event of a data center failure.
- Hub Transport servers should NOT use SAN attached storage for their queue databases. Internal storage, in the Hub Transport server, is preferred. In the event that a CCR lossy failure occurs and Transport Dumpster redelivery is required, it would be best to not have that Transport Dumpster data on the SAN that is possibly out of action now due to a data center failure or SAN failure. If the queue database is located on internal storage in the Hub Transport server, you can simply move that Hub Transport server physically to an operating location (doesn’t even have to be in a data center), put it back on the network and now it can participate in Transport Dumpster redelivery. Cool.
- If possible, Hub Transport servers should NOT be placed on blade server hardware UNLESS you have the capability to place a blade server chassis in another location during a data center failure. This goes along with the previous recommendation about being able to bring that Hub Transport back online in an alternate location.
- At least one Domain Controller SHOULD be placed in each local data center you’ve got Exchange servers placed in. Of course, you should really have a minimum of two data centers to take advantage of CCR and you should always have at least two Domain Controllers per AD site…so one Domain Controller per data center makes perfect sense. If you’re going to use a Domain Controller (as noted previously) for your File Share Witness, this then offers you both AD and Exchange protection in one move.
Do you have anything to add?
Comments
3 Responses to “My “unofficial” Exchange Server 2007 CCR recommendations”
Leave a Reply
-
Translator
[...] Read the rest… [...]
With the Hub Transport Server is it ok to also run the Client Access and Unified Messaging on the same server?
I’d recommend against it, though I didn’t find anything specifically saying one way or the other.
See these links for some general information.
http://technet.microsoft.com/en-us/library/bb232197.aspx
http://technet.microsoft.com/en-us/library/bb123911.aspx