vSphere High Availability

vSphere High Availability (HA) provides easy to use and cost-effective high availability for virtual machines and applications running in virtual machines. HA is part of Cluster and can be enabled from Cluster settings at the time of Cluster creation or later.

HA is a very important and useful feature of vSphere to keep virtual machines up and running with minimum downtime. HA keep eye on all hosts in the cluster and when any host failure happens HA, restarts affected VMs on another host in the cluster.

 VMware HA

Advantages of HA:

VM and services get up with minimum downtime

Auto failover on ESXi hardware failure for all running VMs

Easy to enable and minimal setup

HA gets enabled automatically for newly added host in cluster

No additional cost involved

Prerequisites of HA:

vCenter Server

At least two hosts

At least one Shared Datastore

HA enabled Cluster

How does HA work?

When you enable HA on the cluster it requires at least one shared datastore which must be accessed from all hosts in the cluster. A directory. vSphere-HA gets created on the root of datastore that is used for both datastore hearbeating and persevering the set of protected virtual machines (don’t delete or modify this directory or contents inside it). HA keeps host’s updated state data on the datastore and use datastore for host heartbeats. All hosts in the cluster gets state information of other hosts from this datastore.

Every host in the HA cluster functions as master or slave (subordinate) host. When HA is enabled in the cluster all active hosts participate in the Master election process to become master for the cluster. A host with maximum number of datastores has an advantage in this election. Every HA cluster has only one master and rest of hosts as subordinates. If master fails due to any reason, then again election is held for new master.

When master cannot communicate with any subordinate host over the network then master uses datastore heartbeat to check where host is running or got failed. If subordinate host has stopped datastore heartbeat then master restarts all the VM of the failed host on another host.