Specify a failover host for HA clusters in VMware

Virtualization expert Rick Vanover explains the pros and the cons of using the specified failover host option in VMware's High Availability feature.

VMware vSphere's High Availability (HA) feature allows virtual machines to be restarted on other hosts in the event of a host failure. I have had a love-hate-hate-love-hate relationship with HA throughout the years; I'm keeping score of how many times it has saved me compared to biting me in the rear end.

Putting my mixed feelings about the feature aside, I recently gravitated towards a new configuration option for HA clusters in certain situations. The option to specify a failover host for HA clusters allows a specific ESXi (or ESX) host to be designated as the host to absorb the workload on the failed ESXi host. This option is a property of an HA cluster (Figure A). Figure A

Click the image to enlarge.

This option is set for a test cluster of only two hosts, but some of the attributes are visible quite easily. First, the vesxi4.rwvdev.intra host is designated as the HA failover host; this means that virtual machines are not intended to run on that host in a normal running configuration. This is at the expense of the other host, because there is one extremely busy host and one relatively idle host.

The use of the designated failover host offers the opportunity for administrators to capture some benefits compared to the other HA options. The first option is that you could place a lower-provisioned host in the admission control inventory. This can include using a 2 CPU (socket) host instead of a 4 CPU host that would exist in the rest of the cluster, thus reducing licensing costs. Another benefit is each host that is not the failover host would be allowed to go higher in its utilization, as an admission control policy would not prohibit additional virtual machines on that host.

There are a number of critical decision points on HA, but I would be remiss if I did not mention what I feel to be the authoritative resource for this feature: the HA Deepdive from Duncan Epping's Yellow Bricks blog. Duncan has good information about all of HA, including the designated failover host option.

Probably the best use case for using HA and designating a failover host is to set individual virtual machine HA event response rules. A good example of this would be to not perform an HA failover on development virtual machines, should they be intermixed in a cluster. Figure B shows this configured in an HA cluster where all test and development virtual machines are configured to not have an HA event restart. Figure B

Click the image to enlarge.

This is the proverbial "it depends" configuration item. There are plenty of factors that go into considering this HA cluster arrangement, but the designated failover option doesn't seem to be used that frequently.

Do you use the default configuration? If not, how do you configure VMware HA? Let us know in the discussion.