I have heard two arguments for using HA admission control on DRS clusters when VCAC is used for automated provisioning. (In this case vCloud Director is NOT used)
Pro: Admission control should be used. Because VMs can now be provisioned more quickly, clusters can fill up more quickly, and it is more important to have admission control than before.
Con: Admission control should not be used. The idea of cloud is to create a layer of abstraction between the cloud layer and the virtualization layer. The logic that controls whether a provisioning operation should be approved should therefore all be in the cloud layer (in this case that is in vCAC and Orchestrator). Before VCAC provisions a VM, it should be able to check the capacity of the cluster it is provisioning onto, to determine if there is sufficient space. If a provisioning operation fails, it should fail because it is stopped by the logic in vCAC before it starts, and not because the provisioned VM can't power on after being provisioned, because HA admission control stops it. Instead of HA admission control, part of the provisioning workflow should be when a determination is made of which cluster to put the VM on, to use VCAC/vRO check that cluster for capacity. IF there isn't sufficient capacity, the try/catch/exception logic is handled by vCAC/vRO to handle it. This means that you would never put a VM on a cluster without sufficient capacity only to have it not be able to power on once provisioned.
Can anyone provide more light, arguments, reasoning to add to, refute or otherwise solidify the understanding of how this decision should be handled?