Paul Gregory | 29 January 2013
With Windows Server 2008 R2 and Windows Server 2012, Microsoft extended the capability to produce a Multi-Site/Multi-Subnet/Geo-Cluster. I thought it would be a good idea to address some of the requirements and configuration choices around this solution.
When building a cluster in Windows pre 2008R2,
a clustered resource had a collection of dependencies. For
example if it was a clustered file server, it would have to have a
Name used to connect to it, and IP Address that the name resolved
to and storage for storing the data.
Within previous version of the OS these dependencies were joined with an AND relationship. This meant that all nodes of a cluster had to be on the same subnet, which caused issues and limitations because to achieve a multi-site cluster between your datacentre and your DR Datacentre, a VLAN needed to be stretched.
Today OR's can be used within the relationships therefore allowing either IP address A or IP Address B to come online. It also means these same rules can be attached to the storage removing the requirement for shared storage.
One thing to consider is how network failures will be handled. For example if there is a failure in the network close to the DR site how would you wish the cluster to react? One option would be to introduce a third site.
This third site contains a File Share (highly available if required) which the cluster nodes use within their voting. This way it enable the nodes to work out if the can communicate across your WAN.
No shared storage
Because we do not have shared storage the replication of data between the two locations needs to be addressed. The options are simple but the golden rule is the Windows Cluster engine DOES NOT replicate data.
- Storage replication e.g. EMC RecoverPoint
- Software based replication e.g. Vision Solutions Double Take Availability
- Application Replication e.g. Exchange DAG's or SQL 2012 Availability Groups
This video will demonstrate the configuration and setup of a multi-subnet cluster.