VMware HA Options

[tdubb123]

what does the option under

vmware ha:

Host failures

Number of host failures allowed: default at 1

option do?

[williamarrata]

Customizing HA for Virtual Machines

You can customize HA for restart priority and isolation response:

• Restart Priority determines the order in which virtual machines are restarted upon host failure. Restart priority is always considered, but is especially important in the following cases:

• If you've set host failure to a certain number of hosts (for example, three) and more hosts fail (for example, four)

• If you've turned off strict admission control and have started more virtual machines than HA has been set up to support

• Isolation Response determines what happens when a host in an HA cluster loses its console network connection but continues running. By default, virtual machines are powered off in case of a host isolation incident. They are restarted on a different host, and you usually don't want two instances of the same virtual machine running on two different hosts. You can change that behavior for individual virtual machines.
  
  Disk linking usually prevents two instances of a virtual machine from running on two different hosts, but there may be a problem if you're using NAS or iSCSI storage. In that case, keeping the Isolation Response at Power off (the default) is highly recommended.

To customize HA behavior for individual virtual machines:

1. Select the cluster and click the Edit Settings link, then select Virtual Machine Options under VMware HA.

2. For each virtual machine, you can select from the Restart Priority or Isolation Response pop-up menu to customize its settings.

• Restart Priority indicates relative priority for restarting the virtual machine in case of host failure. Higher priority virtual machines are started first.
  
  Note: This priority applies only on a per-host basis. If multiple hosts fail, VirtualCenter first migrates all virtual machines from the first host in order of priority, then all virtual machines from the second host in order of priority, and so on.

• Isolation Response specifies what the ESX Server host that has lost connection with its cluster should do with running virtual machines. By default, each virtual machine is set to be shut down in the event of a host isolation incident.
  
  You can choose Leave running to indicate that the virtual machine on isolated hosts should continue running even if the host can no longer communicate with other hosts in the cluster.
  
  
  Note: If you add a host to a cluster, all virtual machines in the cluster default to a restart priority of Medium and an isolation response of Shutdown.

Hope that helped.

[WillemB]

If you chose the VMware HA option, the HA screen lets you set the following options:

• Number of Host Failures Allowed - Lets you specify the maximum number of host failures from which you can recover or for which you want to guarantee failover. VirtualCenter allows a maximum of 4.

• Admission Control - Lets you determine whether virtual machines can be started if they violate availability constraints. As a rule, the cluster reserves resources to allow failover for all virtual machines on the specified number of hosts. If you expect that you won't usually power on all virtual machines, you might make admission control less strict. If you do, the cluster icon turns red and failover is no longer guaranteed.

[tdubb123]

>>Lets you specify the maximum number of host failures from which you can recover or for which you want to guarantee failover. VirtualCenter allows a maximum of 4.

Does this mean if I have 5 vms on a single esx host and the esx hosts fails. only 4 (max) can get migrated over to the second esx host?

[tdubb123]

I have a ton of vms. And these are the options:

disabled

low

medium

high

If I have 4 machines that need to be on high, what gets started first?

[MikeAvery]

In cases where VMs have the same effective restart priority, they are started in alphabetical order.

[tdubb123]

is it possible to have the vms not turn off in case of a esx host failure. meaning can the hosts be moved to the second esx host without any vms being turned off?

[MikeAvery]

Not with any currently released Vmware product. Upcoming releases (3.5+) will potentially support realtime host failover capabilities.

HA as it currently exists is a "good enough and at the right price" type of solution. It's better to receive an email notification stating that a box went down and your VMs have been brought up ona different server than it is to jump into the car at 3:00 AM.

[WillemB]

tony,

It means the following.

If you have 10 hosts hosting 50 VM's and you have host failure set to 1 then if 1 host (ESX) fails then all VM's will be restarted on the other host(s).

If you have multiple (e.g. 2 failures) host failures then only the first will have it's VM's resumed. This is to protect your cluster from getting overflowed and laying a big burden on all other systems making your whole enterprise a potential snail.

You have admission control can be set to protect you from overfilling all other hosts. If you want to be able to fail 2 hosts then all other (in my example 😎 hosts should be able to accomodate all VM's of the 2 failed hosts. So setting the number of host failures will cost you resources on all hosts.

Admission control can be set to warn you or to stop you from overfilling hosts.

Hope my explanation makes sense.

P.S. Just to know other options also exist to tell the "failed" host what to do if it is still alive but for example is isolated network wise. I won't elaborate on that at the moment, grasp the concept above first.

P.S. HA brings up the VM's in crash consistent state which means it's like you pulled the power plug out of your pc and put it back in. Be prepared for VM problems.