Provide High Availability for a Server Running Hyper-V
Updated: October 24, 2008
Applies To: Windows Server 2008
In this scenario, the goal is to provide high availability for a server that is running the Hyper-V virtualization feature with Windows Server 2008. By using Hyper-V with the failover clustering feature, you can consolidate multiple servers (as virtual machines) on one physical server without causing that server to become a single point of failure. Instead, in the failover cluster, if the server running virtual machines fails or requires scheduled maintenance, a duplicate server makes the same virtual machines available to clients, through a process known as failover.
When you create a failover cluster where virtual machines running on one physical server can fail over to another physical server as needed, you are clustering the physical servers and virtual machines, not the applications that are running in a given virtual machine. The failure of a physical server (a cluster node) would cause another physical server to take over support of the virtual machines, but the failure of an application within a virtual machine would not. For increased availability beyond what the failover cluster provides, implement close monitoring of the applications as well, so that you can provide a quick response if an application fails.
For best results, for a client-server application that is supported by a virtual machine, if a user performs an action that is interrupted, the client computer should automatically retry the action. This is necessary because a user action might sometimes be performed during a failover, meaning that the action would be interrupted (briefly) while failover occurred and would require an automatic retry to complete without further user action.
For more information, see the following topics:
For example, A. Datum Corporation may want to provide employees at their downtown location with a server that offers applications that run on a variety of operating systems (running as virtual machines) using Hyper-V. A. Datum may want this consolidated server to achieve 99.99% availability (down for less than 1 hour per year). A. Datum is willing to closely monitor applications running on the virtual machines to ensure that they are restarted if necessary, but also wants the ability for the physical server itself to fail over to another physical server if there is a hardware failure or if routine maintenance is required.
For additional deployment goals, Identifying Your Failover Cluster Deployment Goals.