vSphere HA uses admission control to ensure that sufficient resources are reserved for virtual machine recovery when a host fails.
Admission control imposes constraints on resource usage. Any action that might violate these constraints is not permitted. Actions that might be disallowed include the following examples:
Powering on a virtual machine
Migrating a virtual machine
Increasing the CPU or memory reservation of a virtual machine
The basis for vSphere HA admission control is how many host failures your cluster is allowed to tolerate and still guarantee failover. The host failover capacity can be set in three ways:
Cluster resource percentage
Dedicated failover hosts
vSphere HA admission control can be disabled. However, without it you have no assurance that the expected number of virtual machines can be restarted after a failure. Do not permanently disable admission control.
Regardless of the admission control option chosen, a VM resource reduction threshold also exists. You use this setting to specify the percentage of resource degradation to tolerate, but it is not available unless vSphere DRS is enabled.
The resource reduction calculation is checked for both CPU and memory. It considers a virtual machine's reserved memory and memory overload to decide whether to permit it to power on, migrate, or have reservation changes. The actual memory consumed by the virtual machine is not considered in the calculation because the memory reservation does not always correlate with the actual memory usage of the virtual machine. If the actual usage is more than reserved memory, insufficient failover capacity is available, resulting in performance degradation on failover.
Setting a performance reduction threshold enables you to specify the occurrence of a configuration issue. For example:
The default value is 100%, which produces no warnings.
If you reduce the threshold to 0%, a warning is generated when cluster usage exceeds the available capacity.
If you reduce the threshold to 20%, the performance reduction that can be tolerated is calculated as performance reduction = current utilization * 20%. When the current usage minus the performance reduction exceeds the available capacity, a configuration notice is issued.