Using HA with blade servers

[TheVMinator]

If I'm using Blade servers, how should I design placement of HA cluster nodes across different chassis to allow for the failure of the chassis itself?

[Troy_Clavell]

when we use to use blades we would spread our clusters across blade chassis.  For a 10 node cluster we had 3 blade chassis, so in the event of a chassis failure we couldn't lose all our nodes.

[a_p_]

Assuming you are on vSphere 5 you may place as many hosts in a single cluster as you can afford to loose in case of a chassis failure. If you are still running vSphere 4 (or older) you need to take HA's Primary/Secondary model (opposed to Master/Slave with vSphere 5) into account. Take a look at e.g. http://www.cloud-buddy.com/?p=786 which explains this in details.

André

[depping · Accepted solution]

Which version of HA will you be using?

With 5.0 or higher, I would say at a minimum 2 chassis and divide the hosts across the chassis so that if anything happens to 1 chassis you still have 50% of your cluster available.

With 4.1 or lower, ensure you never have more than 4 hosts in a single chassis. HA uses a concept called primary nodes in this release to restart workloads. When your primary nodes fail you cannot restart VMs. There are 5 primary nodes in clusters larger than 4 nodes. This is the maximum you can have. So with 10 nodes in your cluster their will be 5 primary. If all those reside in the same chassis and that chassis fails then HA won't power on your VMs.

[TheVMinator]

The environment is 5.1   Thanks for the input.