Microsoft Clustering on vSphere – Incompatible Device Errors
When setting up a Microsoft Cluster with nodes running in vSphere Virtual Machines across ESXi hosts, I have come across folks who have experienced Incompatible device backing specified for device ‘0’ errors. These are typically a result of the RDM (Raw Device Mapping) setup not being quite right. There can be a couple of reasons for this, as highlighted here.
Different SCSI Controller
On one occasion, the RDM was mapped to the same SCSI controller as the Guest OS boot disk. Once the RDM was moved to its own unique SCSI controller, it resolved the issue. Basically, if the OS disk is configured to use SCSI 0:0, then you cannot put the RDM on SCSI 0:1, or SCSI 0:2. You must put the RDM on SCSI 1:x or SCSI 2:x.
Matching LUN ID
Another reason for the above error is when the RDM is presented to the different ESXi hosts using a different LUN ID. The RDM must be presented to all ESXi hosts (and thus all MSCS nodes) using the same LUN ID.
Another important point for different controller is that you can set the operation mode to physical compatibility mode if clustering across esx hosts. Another over looked point.
A quick query though completely not relevant to this topic.
If there is a requirement to present multiple RDMs to a VM, do they need to be on different SCSI controllers (like 2:0, 3:0 etc.) or can they all be under the same controller..?
Nope. No special requirements for RDM presentation. Only thing to keep in mind is that when sharing RDMs across multiple hosts, make sure you use a consistent presentation LUN ID on all hosts.
Thank you very much for the clarification!