I have a simple setup with 2 Compute Nodes to host 2 VMs: a PROD VM and a TEST VM. The requirement is that if the compute node with the Prod VM fails, the Prod VM should automatically migrate to the second compute node. In order to maximize performance for the Prod VM, there should be no other VMs running on the same compute node. This means that the Test VM should be stopped before the Prod VM is started on the failover compute node.
I want to use 80-90% of the physical memory of the compute node for Prod VM, therefore Prod VM can not be started due to lack of memory if Test VM is not stopped up front.