I’ve had a lot of people ask me about clustering ESX servers. They are usually hoping for something that ensures all their Virtual Machines will stay online in the event of an ESX server going down. I always get the same disappointed look when I tell them it’s not currently possible. Only 1 ESX server can have access to a Virtual Machine’s “disks” at a time.

So what do services like Veritas Clustering offer ESX? These products will provide failover by “restarting” the Virtual Machines that were running on a failed ESX server on a different server. It does not migrate the VM’s like VMotion, or keep them online like a clustered application.

So when ESX1 fails, the VMs that were running on that server go down. Veritas Clustering software restarts them on different ESX servers. Before you discount this technology remember, booting a VM takes a lot less time than booting a physical server, and having it automatically restarted can save a lot of time when things go wrong.

Now, before you run out and buy Veritas, you should know that VMware has released a new feature with their VI3 product called HA (High Availability). HA does exactly what Veritas Clustering does, but even better. HA will allow you to over commit the ESX server and run your VM’s in a degraded mode in the event of a failure. This is for customers who need their systems up, even if they are running slower than usual. For those customers who prefer performance, you can tell HA to only start VM’s is the resources are available for them.

So where does that leave Veritas?