It has been a while since we talked about VM sprawl, but Mike Laverick posted an interesting article on SearchServerVirtualization last week that rekindled our interest in this discussion. His article prompted me to rethink how we look at sprawl since sprawl is not necessarily bad. In fact, the use of the term sprawl is used a little too liberally with a connotation that is too negative.
First, lets talk about bad sprawl - the no-kidding, serious, bad sprawl. This flavor of sprawl comes from virtual machines that are running and no longer doing useful work. Since VMs are easy to deploy, sometimes it is a little too easy to spin up a VM and never use it. Or perhaps a VM is used for testing, then forgotten. Without the standard procurement process of getting a server, provisioning it in the data center, and formally adding it to an asset management system, it is possible to have many of these running in your environment doing no useful work.
This type of sprawl manifests itself as Zombie VMs. The key here is to identify potential Zombies by looking for VMs running near idle with little to no variation in their load. These are potential Zombies since virtual machines can be doing work while running in a low load state. Automated Zombie detection tools, will require that you manually check the list of Zombie candidates before you take the proverbial chainsaw to end the Zombie’s existence.
vOPS Server Standard helps you find these potential Zombies, test them and then eradicate the Zombies while marking the non-Zombies distinctively so you don't have to continuously retest them. No one likes to have to go through blood tests frequently just to prove they are not a Zombie.
So sprawl that results in Zombies is clearly bad. But what about the other type of "sprawl"? Next posting, we discuss how “Sprawl Got a Bad Rap”.