HA and DRS behavior in a 2 node Cluster.
An interesting new DRS behavior was introduced with the latest Update 2 patch that I recently became aware of at a client. If you have a 2 node cluster (Only 2 ESX servers in a cluster) and you have both DRS and HA set up, the HA settings will default to a logical 1 allowed failure state. (Since there are only 2 servers, that is the most and least amount of failures the cluster can accept).
This behavior has always been the case however, in the past, when you have entered a HOST into Maintenance mode, DRS would automatically vMotion the running Virtual Machines to the secondary host provided you had set DRS to fully automated. This is no longer the case. The new DRS logic will no longer vMotion running Virtual Machines to a secondary host in a 2 node cluster since entering maintenance mode would cause the HA cluster to violate its ‘1 allowed failure state’.
You can still manually vMotion the machines off and once clear of running machines, the host will enter maintenance mode but DRS will no longer do it for you. Quite annoying when you CHOOSE to put a machine into maintenance mode.
You can read more about it @ http://communities.vmware.com/thread/122653?tstart=0&start=30