An administrator determines that when a host residing in a High Availability (HA) cluster fails, the virtual machines fail to restart on a remaining host, or will only restart after a long delay. Whic

vSphere HA only monitors and restarts virtual machines that are in a protected state at the moment the host failure occurs. If a VM was marked as unprotected - due to admission control, a recent HA reconfiguration, or the VM being added before HA could establish protection - it will be excluded from the automatic restart process entirely.

DRS is responsible for load balancing and initial VM placement but is not required for vSphere HA to detect a host failure and restart VMs - HA operates independently of DRS.

When a host fails, HA must place the affected VMs on the surviving hosts. If those hosts do not have enough free CPU or memory to accommodate the restarting VMs, HA will either fail the restart or queue it until resources become available, resulting in no restart or a significant delay.

Physical Raw Device Mappings do not prevent vSphere HA from restarting a VM; as long as the mapped LUN is accessible from the destination host, HA can restart the VM normally.