Site resilience is a requirement of a lot of companies but on a lot of occasions it is not a requirement to achieve full local resilience immediately following a site failure. In other words a lot of companies are deciding that in the event of the loss of a single data centre site the service would resume in a second data centre but using only a fraction of the kit and might run the core services only. After the service has resumed a decision would need to be made as to whether to establish full service and resilience in the 2nd data centre or choose to fail back…
So the easiest way to achieve this with Exchange 2007 and the mailbox role is to have your SCR target as a single server that could be built out as a cluster in the event of site failure. So is this technically possible and supported? The answer to both of these questions is yes …but with some caveats.
- The SCR target should be built as a single node using a local quorum and should not be built as a Majority Node Set (MNS) cluster with a File Share Witness (FSW). Of course at some point in the future the second node would be added and at that point the cluster would be reconfigured as a MNS cluster with a FSW.
- In order to activate the single node as a SCR target the passive mailbox role needs to be installed and not the active mailbox role. If you install the active mailbox role then you then have a CMS. As soon as a CMS exists a server cannot be a SCR target.
There is now a lot of great information on high availability in general here: High Availability
More specifically this section (Standby Continuous Replication: Site Resilience with Standby Clustering) provides a run through of the processes involved in activating an SCR target. The steps are based on Exchange 2007 running on Windows 2008 but there are references to Windows 2003 where relevant.
Of course, as this section of Technet does touch on, ‘High Availability’ should apply to more than just the mailbox role and in my experience designing to reach your required levels of high availability for client accessibility, message routing etc etc. can be just as challenging…