Requirements and Recommendations for Failover Clusters in Which All Nodes Run Hyper-V

  • Article

Applies To: Windows Server 2008

This topic provides information about requirements and recommendations for a failover cluster in which all nodes (servers) run Windows Server 2008 and Hyper-V. For a description and diagram of the design that these requirements and recommendations apply to, see Design for a Failover Cluster in Which All Nodes Run Hyper-V.

The following list provides information about requirements and recommendations for a failover cluster in which all nodes run Hyper-V:

  • Apply the hotfix (strongly recommended): We strongly recommend installing the hotfix described in the Knowledge Base article at https://go.microsoft.com/fwlink/?LinkId=125397. With the hotfix, the configuration and management of virtual machines in a failover cluster are simplified.

  • Use disks in clustered storage: When choosing the disks to use for clustered virtual machines, choose only disks (LUNs) in cluster storage (listed as Available Storage in the failover cluster snap-in). This ensures that no matter which node owns the clustered virtual machine at any given time, any file that is needed by the virtual machine is accessible to that node. Do not place a file that is needed by a clustered virtual machine on the hard disk of an individual cluster node.

  • Use a separate disk for each virtual machine: We strongly recommend that for each virtual machine, you use a separate disk (LUN) in cluster storage. This is clearly a requirement if you configure a virtual machine to attach directly to a physical disk (sometimes referred to as a pass-through disk), because the virtual machine cannot use the same disk as another virtual machine. However, it is also important in situations where virtual machines use virtual hard disk (VHD) files: the VHD file or files for one virtual machine should not be located on the same disk in storage as the VHD file or files for another virtual machine. Using a separate disk ensures that virtual machines can fail over independently of each other, rather than having to fail over together.

  • For large numbers of virtual machines, identify disks by GUID: A failover cluster with large numbers of virtual machines is like any cluster with a large number of clustered instances requiring a large number of disks. In such a cluster, identifying disks by drive letter could cause you to run out of letters, so identifying disks by GUID is the best option. Note that with the hotfix, this option is easier to deploy than previously, because it is now easier to copy disk GUIDs as needed during the configuration process.

    For disks that are used by virtual machines, the use of mount points is not recommended because the volumes that are mounted on the same drive must fail over together along with the drive that they are mounted on. This would mean that all virtual machines using volumes mounted on the same drive would have to fail over together each time a failover is required for any of the virtual machines.

  • Understand and plan for the resources that must fail over when a given virtual machine fails over: If a virtual machine requires multiple disks, all the disks must fail over together. Keep this in mind as you plan.

  • Maintain records of disks used by virtual machines: After you create your configuration, it can be useful to store disk identifier information (such as disk GUIDs) in a place where you can easily refer to it, for tracking purposes.

Additional references