Hi Prasanth,
Here is what happens when you turn on HA,
1. As soon as you turn on HA, VCenter Server pushes and installs an agent in all the hosts in the cluster.
2. This Agent is known as FDM agent (Fault Domain manager). This is the actual agent which participates in a cluster election process and sends and receives heartbeats.
3. Once the VCenter server installs the FDM agent, VCenter's role is over as far as the cluster's HA functionality is concerned.
4. Now these FDM agents, participates in an election process and after the election, one of the FDM agent becomes a "Master" and others becomes "Slaves".
5. Since these FDM agents sits on the hosts, we call the winner of the election as the Master host and rest of the hosts in the cluster are known as Slave hosts.
6. The election process is simple, the host which can see maximum number of datastores wins the election and becomes the master. In case there are more than one host which can see the maximum number of datastores (Lets say in case of a Tie where two or three hosts can see the same number of maximum datastores) the host, having lexically greatest MOID (Managed Object IDentifier) value wins the election. So in case of a Tie during this election process, lets say the MOID value of three hosts which can see maximum number of datastores are 98, 99 and 100. The host with MOID 99 wins the election as this is the meaning of lexically greatest. they start the comparison from the first digit, so in case of 99 and 100, since 9 is greater than 1, it wins. Similarly in case of 98 and 99, since the first digit are same, it compares the next digit and again since 9 is greater then 8, host with 99 MOID wins.
7. So once the HA is configured and this election process is over, all the hosts FDM agents send and receive heartbeats to the master host FDM agent. In case one of the host fails to send the heartbeat, the Master Host, checks the datastore if there is any lock file for the failed host, if it finds none, the master host, initiates the restart of the VMs on the failed host.
8. If the Master host itself fails, then the remaining FDM agents undergo a fresh election process again and a new Master is elected and this new master initiates the restart of the VMs from the failed host to other remaining hosts.
I hope it helps, please get back to me if you have any more doubts.
