What is the ideal desgin for primary node distribution for HA in a cluster. I came across a question that read " be able to design or build highly redudant HA config where the primary nodes are spread out". I know in HA u have 5 primary nodes and the rest are secondary and if you loose all 5 primary nodes you are done!
In the event of hardware failure you do not want all of your HA primaries to fail.
So if you have a 16 node cluster with 8 blades in each enclosure you would want HA primaries in each enclosure.
If you have rack mount servers with redundant storage and networking then it doesn't really matter.
Look at your design and distribute the hosts carefully. Remember for the first time when you start adding your hosts to the cluster it will be treated as first come first serve.
So if you have 2 full house Chasis (32 nodes) then carefully add the hosts into cluster one by one. I would suggest first add 3 nodes from one chasis and then add the other two from other chasis.
So in this case even if one of your chasis fails, you will still be benefitted by HA as other nodes will be able to restart those VMs (depend on % available resources).
There was also a VMworld 2010 session about this kind of design.