Title: HA datastore heartbeats prevent host isolation

Description: What are datastore heartbeats? The host-X-hb (where X is the host’s MOID) is Located on each heartbeat datastore, this file is used to check for slave liveness through the heartbeat datastore. This file is checked by the master host if the master loses network heartbeats from the slave. For VMFS datastores, the vSphere HA agent locks this file with an exclusive lock and relies on the VMkernel heartbeat to indicate liveness. For NFS datastores, vSphere HA periodically updates the time stamp to this file to indicate liveness. The host-X-poweron (where X is the host’s MOID) is a per-host file that contains the list of all virtual machines that are powered on. This file is used as a communication channel if a management network outage occurs. Isolated slaves use this file to tell the master that it is isolated as well as to tell the master which virtual machines it has powered off. The master host must determine whether the slave host: • Actually crashed • Is not responding because of a network failure • The HA agent is in an unreachable state The absence of both a network and datastore heartbeat indicates full host failure. Datastores are used as a backup communication channel to detect virtual machine and host heartbeats. Datastore heartbeats are used to make the distinction between a failed, an isolated or a partitioned host.

Date: 12/11/2011 12:00 AM

Author: Eric Sloof