HA(High Availabilty) and Vcenter server Reboot.

I do not know what they mean. will the VMs be rebooted or they mean moved( when they say Restarted on the other host) ?

This is correct, because VMware vCenter Server ONLY configures VMware HA, VMware HA is controlled by Agents which are installed on the ESXi hosts, and perform Master and Slave elections, it's the Master and Slave Agents which perform VMware HA.

The HA Agents on the hosts...

vSphere HA State - Master - A server which is elected as the master. This agent monitors the VMs on this server, and other operational Hosts, and it WILL attempt to restart VMs on failure.

vSphere HA State - Slave - This server is connected to the Master Agent, via the Management Network. The vSphere HA Protected VMs on this server are monitored by one or more vSphere HA Master Agents, and the agent will attempt to restart VMs after a failure.

vSphere HA Protected VM - vSphere will attempt to restart the VM after a supported failure of the VM.

VM is HA Protected on the following conditions:-

VM is in a vSphere HA enabled cluster.
VM is powered on successfully after a successful user power on.
vSphere HA has recorded that the power state is ON.

When an ESXi Host Server Fails (which is part of a VMware HA Cluster), all the Virtual Machines, which are hosted on that Host, also go down, e.g. fail.

A Host Failure could be:-

1. Pink/Purple Screen of Death - caused by memory fault.
2. Pink/Purple Screen of Death  - cause by cpu fault.
3. Power supply failure (if only a single power supply)

A Host Manual Shutdown, reboot, restart is not considered a host failure. Because it's a controlled shutdown.

So we have a Host which has failed, and ALL the VMs it was hosting are now DOWN!


When a host server fails, all the VMs which are on that host server will also FAIL. So a Host fails, all the VMs would go OFF (crash!). They cannot be Live Migrated or vMotiond or MOVED in advance, because how would the server know it's going to fail - it cannot!

So VMware HA, issues a restart of failed VMs, on the remaining hosts!

So, yes, the VMs are restarted, and yes the VMs are moved, because the host they were running on has failed, and no longer available in the cluster.

VMs running on new hosts, BUT they are not MOVED LIVE!!!! No vMotion! They do not really move anywhere, VMs are stored on shared storage, they do not move, but the process is executed on a different host!

So it depends on how you regard "move". They have moved, because they could of been on Host A, and now they are on Host B or Host C, or Host D but also Host A has gone! So they could no longer be running on Host A.

So, expect VMs to be restarted within 1-2minutes of a host failure.

If you are a VMware Admin, I would recommend Duncan's Books on Clustering here

vSphere 5.1 Clustering Deepdive

Any questions?

hanccocka:

My concern was about the Vcenter.
I have restarted the Vcenter server, and in our Cluster HA and DRS are enabled.
I want just to make sure that when restarting Vcenter server , the VMs will not restart in other host as the article described it (unless if I misunderstood the article).

For the Failure of the host , I amaware of that,  HA will reboot the VMs in the other host within the cluster.

**Recently I put ESX host in maintenance mode, some VMs moved automatically and some did not move at all. I had to move them manually. Some VMs were turned off and as you mentioned in a different post "make sure  CD/DVD drive:1 not connected "  , I saw the VMs that were turned off showing  "Connected to ISO image on local disk".

So , I had to shut down the host with the VMs that were turned off still displayed under the host.

Excellent hanccocka !!!