Using CCR for site recoverability

The ability to be able to continue providing a full service to your user community in the event of the loss of a datacentre is an increasingly common requirement.  The use of CCR with Exchange 2007 is an obvious choice in going some way to meeting this requirement. If you have the ability to stretch a subnet between your data centres then a Windows 2003 cluster with CCR is an option.  CCR does not guarantee complete recovery of all data in all scenarios but may be considerably cheaper than storage replication solutions and does mean that a disaster recovery scenario is managed from within one team rather than several.  In most cases the Exchange team can manage the entire recovery in the event of data corruption rather than rely on intervention from your storage team for example.

If you are considering this solution then I would definitely recommend testing this in a lab environment to make sure that you understand exactly what is going to happen in each scenario you plan to be able to recover from.  For example if data corruption means that a single storage group is unavailable - what should your recovery strategy be? Do you failover your entire server with potentially 50 storage groups in order to restore the service for this single storage group.  This means the temporary loss of service to all users for the sake of those on one mailbox store.  It is also essential that you understand how to recover a CCR cluster in the event of the loss of a single site.  It may not be just a matter of running a powershell commandlet.  A MNS cluster requires a File Share Witness.  If the FSW is in the site that has been lost then this will impact your recovery paths. The following blog discusses this particular issue in more detail. (

With Service Pack 1 SCR (Standby Continuous Replication) will be introduced which again offers another, perhaps more favourable option, if site recovery is your requirement.  There are already some great blogs and articles on CCR and high availability.  Have a look at these:

 ...more on this soon.