A Prism Central admin is sizing AHV hosts for a VM cluster that must survive a host failure. What should be configured so placement decisions account for failover capacity?

AHV VM High Availability restarts VMs from a failed host on the surviving hosts. In its default best-effort mode, HA restarts as many VMs as free resources allow, with no promise that every VM will fit. Enabling HA with resource reservation, called Guarantee mode, makes the cluster set aside capacity across the hosts: one host's worth of resources when all containers are RF2, or two hosts' worth when any container is RF3. AHV then accounts for that reserved capacity when placing and powering on VMs, refusing placements that would eat into the failover headroom, so a host failure can always be absorbed. Memory overcommit works against this goal; it squeezes more VMs onto hosts by reclaiming unused guest memory, reducing real headroom rather than protecting it. Best-effort HA is the default behavior and gives no guarantee, so it fails the requirement. Anti-affinity rules spread VMs across hosts so one failure doesn't take down every instance of a service, but they do not reserve capacity on the remaining hosts. Exam caveat: AHV has no per-VM CPU or memory reservations, shares or limits; capacity guarantees for failover come from HA Guarantee mode. Operational check: after enabling it, review the HA reservation and the remaining memory available for new VMs in Prism before adding workloads.