Clustering Virtual Machines in Production
Yesterday (5/19/09), the KB article which details the support policy for SQL Server failover clusters was officially updated to include support for clustering guests/Virtual Machines … including in production. So anyone who talked to me at TechEd last week, I was right until yesterday 🙂
Now a couple of points which the KB does not say currently (and I hope it will):
1. Don’t put both of your guests under the same hypervisor. It makes no sense, and creates a single point of failure for both of your cluster nodes. Nothing will stop you from doing this since both of your guests have no idea they are virtual machines.
2. Don’t mix physical and virtual/guest nodes. Choose a single strategy. There is nothing that will prevent you from mixing node types, but it really makes no sense to me.
Remember that each of the virtual/guest nodes running under the hypervisor of your choice must also meet the requirements for Windows failover clustering as if they were nodes on physical hardware. That means they must have redundant everything, which goes back to my point #1.
Other helpful links which apply:
KB943984 The Microsoft Support Policy for Windows Server 2008 Failover Clusters
KB309395 The Microsoft support policy for server clusters, the Hardware Compatibility List, and the Windows Server Catalog (pre-Windows Server 2008)
KB327518 The Microsoft SQL Server support policy for Microsoft Clustering
This is all good timing as I am finishing up my book. It does make things a bit difficult as some chapters are done, but this will make it in one way or another.
It was a good week. Thanks to everyone who came to the booth and to my session. I do look and incorporate feedback from the sessions I do, so anyone who posted evaluations, it’s helpful.
The Tech Talk I did (Clustering Microsoft SQL Server 2008 with Windows Server 2008) is also posted, so have a look. It has some demos and is about 20 minutes long give or take.