I wanted to take some time to focus on what has changed with Admission Control since vSphere 6, as this is significant and has even received roomfuls of clapping from customers in our betas and on-sites, apparently we’ve fixed a big pain-point for the customers :).
When you click on the Admission Control tab you will see the image below. What’s unique about the 6.5 release is that the resources required for failover capacity is now tied directly to the Host failures cluster tolerates box. What I mean is that, say we have 4 hosts in a cluster. If I change the Host failures cluster tolerates to ‘2’, it will automatically set CPU and Memory to 50% failover capacity. Likewise if I switch it to ‘1’, it will automatically set CPU and Memory to 25% failover capacity. No more having to manually set the failover capacities unless you want to (see the ‘Override calculated failover capacity’ checkbox).
Additionally, the feature that most folks are jumping for joy about, is that the default Admission Control policy has changed from Slot Policy (Default until 6.5), to ‘Cluster Resource Percentage’. We had found that very few people were actually using slot policy, and those that were, most likely weren’t using it correctly. We’ve updated this now and moving forward it will be the default policy.
But wait, it gets better! Not only is that the default, but using the Host Failures Cluster Tolerates as the basis for Admission Control failover capacity, any time you add or remove a host from the cluster, the failover capacity percentages will update here, and the amount of resources required on each host will also be updated automatically. NO MORE NEEDING TO REMEMBER TO GO IN AND UPDATE THESE SETTINGS WHEN A HOST ENTERS OR LEAVES THE CLUSTER!
The final item I wanted to touch on is the Performance degradation VMs tolerate. This is a new setting that will allow you to change when a warning is raised on insufficient failover capacity. Let’s say you are overcommitting and have many more VMs in your cluster than you really should (from an HA perspective). You can now set an amount of acceptable performance degradation to allow for more VMs to power-on before the alert/warning appears, allowing you to trade off a % of performance for more availability.