Procedure to relocate shared RDM pointer files from one datastore to another

Products

• VMware vSphere ESXi

Issue/Introduction

In a Windows Failover Cluster (WSFC/MSCS) environment using shared Raw Device Mappings (RDMs), the pointer files (.vmdk) may reside on an incorrect or non-preferred datastore.

 

Environment

VMware vSphere 7.x

VMware vSphere 8.x

 

Cause

Because these RDMs are shared across multiple cluster nodes and utilize SCSI-3 Persistent Reservations, they cannot be moved using standard live Storage vMotion. Attempting a live move will typically fail or cause cluster instability.

 

Resolution

• Record the LUN ID and naa ID for each RDM to ensure the correct physical device is re-added later.
• Gracefully shut down all Windows nodes (VMs) that are members of the clustering solution.
• On all nodes, edit the VM settings and remove the RDM hard disks.

Note: When removing from the first/secondary node, select "Delete files from datastore" to eliminate the pointer files from the incorrect datastore. This action only deletes the small metadata pointer file, not the data on the physical LUN.

For subsequent nodes, simply remove the disks from the configuration.

• Select the primary VM node > Edit settings > Add New Device > RDM Disk.
• Select the appropriate RDM LUN.
• When prompted for the location of the pointer file, select the correct/target datastore.
• Ensure the SCSI ID matches the original configuration and the Bus Sharing mode is set correctly according to the RDM mode (Physical/Virtual).
• Select the remaining VM nodes and go to Add New Device > Existing Hard Disk.
• Browse to the target datastore and select the pointer file.
• Power on the cluster nodes one at a time.