Evidian SafeKit implements Hyper-V replication automatic failover and load balancing between 2 nodes. This article explores how to combine a Hyper cluster with an application cluster in order to implement a cluster on 3 servers with real-time synchronous replication and automatic failover.
Note that Evidian SafeKit implements Windows file replication and Windows clustering between 3 servers without the need of Hyper-V. The advantage of the Hyper-V synchronous replication cluster is that the failover and the failback to the 3rd server is automatic. There is no manual procedure to apply in the SafeKit web console.
Note also that SafeKit implements Hyper-V synchronous replication and not the traditional Hyper-V asynchronous replication. This feature is important because with an asynchronous replication, there is data loss on a failover as explained in this article: synchronous replication vs asynhronous replication.
To implement the solution, 2 SafeKit mirror modules are required:
If server1 fails, the Hyper-V mirror module restarts VM1 on server 3. The application cluster implemented inside VM1 and VM2 now runs between server 2 and server 3. If there is another failure of server 2 or server 3, the application module will manage it and will restart the critical application.
When server1 comes back, the Hyper-V mirror module resynchronizes automatically VM1 on server 1. The resynchronization is made without stopping VM1 on server 3. This failback operation is automatic and does not require any manual operation.
To return to step 1 with VM1 running on server1, either the operator clicks on the swap button of the Hyper-V module in the SafeKit web console. Or by configuration of the Hyper-V module, the swap of VM1 on server1 is automatic.