When the primary host in a vSphere HA cluster can no longer communicate with a subordinate host over the management network, the primary host uses datastore heartbeating to determine if the subordinate host might have failed or is in a network partition. If the subordinate host has stopped datastore heartbeating, that host is considered to have failed and its virtual machines are restarted elsewhere.
vCenter Server automatically selects a preferred set of datastores for heartbeating. This selection is made with the goal of maximizing the number of hosts that have access to a given datastore and minimizing the likelihood that the selected datastores are backed by the same storage array or NFS server. In most cases, this selection should not be changed. To see which datastores vSphere HA has selected for use, in the vSphere Web Client you can go to the cluster's Monitor tab and select vSphere HA and Heartbeat. Only datastores mounted by at least two hosts are available here.