A. My last Q&A explained the first of HA's admission control policies. That one, Host failures cluster tolerates, sets aside a quantity of resources that's equal to the cluster's most powerful host. Most experts consider it the best policy for clusters that are comprised of a sufficiently large number of hosts, because as your cluster scales, admission control always rebalances itself to ensure a certain number of host failures can be protected.
But some clusters aren't that big. Maybe you can't afford ten ESX hosts for your cluster—only three or four. At the same time, you don't want to set aside a full host's contribution for admission control. That's a lot of waste for your hard-earned hardware dollar!
Another of my recent Q&As discussed that a course of action in this situation is to simply disable admission control. If you don't care that a host failure and its subsequent relocation of virtual machines (VMs) might affect the performance of every other VM, then disabling admission control isn't necessarily a bad idea.
Most IT pros want VM availability and good performance, so disabling admission control won't work. Instead of losing performance everywhere in the case of a failure, you might be willing to deal with the fact that some VMs won't be allowed to power on.
In this situation (and some others), choosing the admission control policy Percentage of cluster resources reserved as failover capacity might be smart. When you select this policy, you're given the option not to select a number of hosts. Instead, you specify a percentage of total cluster resources that remain in reserve.
With this policy, it becomes possible to set a smaller percentage of reserve than the per-host policy is limited to. For example, in a four host cluster you might not want a full 25 percent of resources—an amount equal to one host's contribution—to be set aside. You might be willing to set aside 10 percent, though.
Doing this obviously means that some VMs won't be allowed to power on. Those VMs, as admission control confusingly states, violate availability constraints.
This setting isn't the best answer. The best answer is to simply buy more hardware. But when budgets are tight, most IT pros are forced to make do with what they have. This second policy gives you the option to do that.
Coming up: Want to prioritize which VMs get restarted and which ones don't? I'll talk more about it in my next Q&A.
Catch up with @ConcentratdGreg on Twitter!
