HOWTO: Migrate Disks of a Multi-node Microsoft Windows Cluster Workload without Downtime

This procedure is for zero-downtime migration of a set of disks of a multi-node Windows Failover Cluster. Perform the procedure for each set of disks that are owned by a particular node where CMC agent is deployed. For multiple sets of disks owned by different nodes, perform the same procedure at each owner-node. For preparation of the migration, the new destination storage can be prepared and tested to ensure the new storage is capable of presenting new storage to each cluster node. However, if new disks are pre-created as migration destination, do not present them to all nodes prior to migration. Only present the new disks to each of the owner node of the disk set to be migrated. However, if the new storage can only present disks to all nodes and not individual nodes, then use Disk Manager at each node to verify the new disks are “offline” for all nodes including the owner node.

Procedure

Deploy CMC on the node that owns the disks to be migrated. Going forward, we'll refer to this as the CMC Node.

Ensure the clustered disk being migrated is only active on the CMC Node.

Ensure the destination disks are only visible to the CMC Node and has not been LUN masked to any other node. If the new storage has no way to LUN mask disks to individual nodes of a cluster and must LUN mask to all nodes, then ensure each of the other nodes keeps the new disks “Offline” using Disk Manager.

For any disk that has been configured as Clustered Shared Volume, set the mode to “Redirected Access.” In this mode, only the CMC Node can write to the disk. Verify the CMC Node is the owner.

Create migration session, then wait for the migration to complete and enter Tracking mode (periodic resync mode). The system can remain in this state until it is time for cutover. If any of the disks being migrated changed ownership to other cluster nodes for any reasons, the Migration Session will be in a Failed state and needs to be start from the beginning by deleting the entire Migration Session and this procedure repeated from the beginning.

When cutover time arrives, remove the source disks from all other nodes by removing the FC zone and/or unmasking the LUNs at the source storage. The CMC Node should be the only node that has access to the source disks at this point.

At all other nodes, open Disk Manager and keep performing a rescan until the source disks disappear.

Verify the destination storage is already zoned with all nodes and present (LUN mask) the migrated destination disks to all nodes, then rescan. The new disks will appear in Disk Manager of each node but will be Offline and appear to be “Uninitialized.” Keep them offline. For Windows Disk Manager, an Uninitialized disk has a black frame.

At the CMC Migration Session, trigger cMotion, confirm the state of the Session is “cMotion”. Then, trigger Finalize Cutover to activate the new disks.

At all other nodes, rescan and confirm the new disks are no longer “Uninitialized.” They should remain Offline, but the frame of the disks in Disk Manager became blue in color, indicating that a valid partition exists.

Using Cluster Manager, move the disks from the CMC Node to the other node(s).

Using Cluster Manager, Pause the CMC Node to put it offline so that it will not be able to take back ownership of the disk.

Confirm on the new owner node, using Disk Manager, that the volume labels of the new disks are now visible and that the new disks are online automatically.

Delete the Migration Session so that cMotion redirection is no longer in effect.

Remove the source disks from the CMC Node and rescan to ensure the source disks disappeared. Keep the destination disks offline.

Use Cluster Manager to bring the CMC Node online, and “move” all the new disks back from all nodes to ensure proper cluster disk movement.

Redistribute the new disks to all nodes. Migration is complete.

For Cluster Shared Volumes, turn off the “Redirected Access” mode if so desired.
Was this article helpful?
Cancel
Thank you!