We've got a 4 node NCS cluster running entirely virtualized on VMware ESX. Each of the 4 nodes lives on a seperate ESX host, with RDM mappings back to data on our SAN. This was basically created using the same steps in Cool Solutions: Configuring Novell Clustering with ESX Server

This works quite well, however, one of the major drawbacks of the above is the lack of vmotion-ability. Now I understand two nodes can't live on the same ESX host because of the RDMs (and thus would be attempting to access the same RDM mappings). However, we have more than just the 4 ESX hosts and would like to be able to vmotion any NCS cluster node to another ESX server not currently hosting any of the other NCS nodes.

The limiting factor seems to be the SCSI Bus Sharing setting on the SCSI controllers being set to physical (as recommended, with little/no explaination, by the cool solutions article). Now, I'm no VMware storage expert, and based on the description of the setting, this seems to make since. However, another staff member here mistakenly setup nodes and left this at the default value of 'None' and the node seemed to function just fine.

So my question is, in the above scenerio, what are the implications to leaving the SCSI Bus Sharing at the 'None' default value (against the directions of the cool solutions article) in an effect to mantain vmotion-ability? is it possible to have an entirely VMware virtualized NCS cluster using RDM's and still have vmotion capability?