Datastore Expansion Failure on Multi-Extent VMFS-6 Datastores
search cancel

Datastore Expansion Failure on Multi-Extent VMFS-6 Datastores

book

Article ID: 418743

calendar_today

Updated On:

Products

VMware vSphere ESXi

Issue/Introduction

Symptoms : 

1. Datastore expansion fails on multi-extent VMFS-6 datastores with the error: "Operation failed! Cannot change the host configuration."
2. The datastore status displays as Offline in the vSphere Client.
3. vCenter inventory reports incorrect or stale capacity/size information for the datastore.
4. Resident virtual machines may experience low disk space conditions or performance issues.
5. This issue occurs even after the underlying LUN space is increased on the storage array.

 

Environment

VMware ESXi 8.x
Multi-extent VMFS-6 datastores backed by Fiber Channel or iSCSI LUNs.

Cause

The issue arises from architectural constraints of multi-extent VMFS volumes combined with incorrect expansion procedures or array connectivity issues:

  1. Expansion Constraint: VMFS-6 requires expansion to occur at the tail extent. If space is added to a non-tail (head) extent, the unallocated space cannot be utilized because VMFS concatenates extents sequentially.
  2. Offline Status: If the underlying LUNs backing any extent in the multi-extent datastore become disconnected or offline at the storage array level, the entire datastore will be marked Offline and metadata updates (including expansion) will fail.

Resolution

If the issue persists due to a non-contiguous expansion (expanding a non-tail LUN), follow the steps below to migrate to a stable single-extent configuration:

  1. Verify Storage State: Ensure all LUNs backing the datastore extents are online and mapped to the ESXi hosts at the storage array level.
  2. Provision New Storage: Request the storage team provision a single, new LUN of sufficient size to hold all data from the multi-extent datastore.
  3. Create New Datastore: Format the new LUN as a single-extent VMFS-6 datastore.
  4. Migrate Workloads: Perform a Storage vMotion migration to move all virtual machines from the multi-extent datastore to the new single-extent datastore.
  5. Decommission Old Datastore:
    • Unmount the old multi-extent datastore from all hosts in the cluster.
    • Delete the old datastore.
    • Reclaim the old LUNs at the storage array level.

Additional Information

To prevent this issue in the future, always expand the tail LUN in a multi-extent configuration. However, using single-extent datastores is the Broadcom recommended best practice for simplified management and expansion.