Scott's Weblog The weblog of an IT pro focusing on cloud computing, Kubernetes, Linux, containers, and networking

Distance vMotion = Stretched Cluster?

Apparently, some people within the virtualization and storage community took umbrage with one of my statements from my article on stretched clusters and distance vMotion. Specifically, this statement:

Long-distance vMotion and stretched clusters are not the same thing.

I’ll be the first to admit that I don’t know everything, and I try (not always successfully) to be open to admitting where I’m wrong. So feel free to (constructively) tell me where I’m mistaken here. How can long-distance vMotion and stretched clusters be considered the same?

Yes, both of them require Layer 2 adjacency for VMs. Yes, both of them need an appropriately configured storage layer underneath them (the architecture of which is an entirely separate discussion, in my opinion—or is it?) so that storage is accessible from both sites. But the fact that they share common requirements doesn’t make them the same. Related? Yes. The same? No.

We know that a cluster is not a vMotion boundary. In other words, you can use vMotion to move VMs between clusters. That means I can have two clusters—one in each site—and I can use distance vMotion to move VMs between them. Ergo, distance vMotion but not a stretched cluster.

Similarly, I could have a stretched cluster (a cluster of ESX/ESXi hosts with some hosts in one site and some hosts in another site) and not use distance vMotion, since a cluster does not automatically mean vMotion. (Thnik about a cluster with DRS disabled.) Ergo, a stretched cluster but not distance vMotion.

So what am I missing? Or am I just being too technical with the definitions? Constructive comments and feedback are welcome.

Metadata and Navigation

Be social and share this post!