vSphere : Building a Windows Cluster
Clustering Windows Server 2003, or 2008 under vSphere is a simple process:
- Create your two VM cluster nodes in the vSphere Console
- Install Windows Server Enterprise Edition on both nodes
- Add additional ‘cluster’ storage, attached to an additional shared SCSI controller (for Single Copy Clusters, SQL clusters or File/Print Clusters then do not use Paravirtual, use LSI Logic. Paravirtual is fine for an Exchange 2010 DAG.). You’ll get a new controller if you assign a SCSI ID to each new disk starting ‘1:’
- Ensure all disks that are shared are set to Independent Persistent
- Ensure all Disks are “eagerZeroedThick”
- Ensure the controller is set to Virtual Sharing in the VM configuration
- Build the cluster
The step that is often missed out is the modification of the vmx file for each cluster VM node; without these changes you’ll find that NTFS filesystems can become corrupted or SQL databases end up with torn pages / in a suspect state. You may even see chkdsk running after a failover from one node to another.
Add the following lines to the bottom of your vmx file to help alleviate these symptoms:
{code lang:ini showtitle:false lines:false hidden:false}disk.locking = “FALSE”
diskLib.dataCacheMaxSize = “0”{/code}