VMware ESXi and vSphere Cluster Management

How to Unmount a VMFS Datastore from an ESXi Host

Learn how to safely unmount a VMFS datastore from selected ESXi hosts in the vSphere Client without deleting its files or formatting its storage.

Unmounting a VMFS datastore removes an ESXi host's active access to an existing datastore. It does not erase the VMFS filesystem, virtual machines, virtual disks, templates, ISO files, or other data on the underlying storage.

This operation is useful before storage maintenance, host decommissioning, or removal of storage presentation. Perform the safety checks first, then unmount the datastore from the exact hosts identified in your change plan.

Key storage terms

VMFS is the VMware Virtual Machine File System used by ESXi datastores. A datastore is a logical storage container that ESXi uses for virtual machine files and related objects.

Unmount means removing one ESXi host's active access to an existing datastore while preserving its contents. A shared datastore is presented to, and potentially mounted by, more than one ESXi host. A LUN is a logical storage device presented to an ESXi host and often used as the backing device for a VMFS datastore.

What unmounting does and does not do

Unmount datastore
Removes access from the selected ESXi host or hosts.
VMFS and its files remain intact.
Commonly used before maintenance or host retirement.

Mount existing datastore
Restores host access to an existing VMFS datastore.
Does not require creating or formatting a new datastore.

Detach storage device
Removes an ESXi host's use of the backing device after datastore access has been removed.
This is a separate, more advanced storage operation.

Delete datastore
Removes the datastore configuration and is intended to remove the datastore from the storage device.
Use only when data removal is explicitly planned.

Format or create a datastore
Creates a new filesystem and can destroy existing data on the selected device.
Never use this to restore an unmounted datastore.

For example, if a shared VMFS datastore is mounted on three hosts and you unmount it only from one retiring host, the other two hosts can continue to see and use it. The datastore remains available wherever it is still mounted.

Pre-unmount safety checks

Record the datastore name, its backing device or LUN identity, the affected hosts, and the intended maintenance activity. Confirm whether the storage is local to one host or shared among several hosts, and identify exactly which hosts should lose access.

No virtual machines or required files in use
Running or registered virtual machines, configuration files, virtual disks, snapshots, templates, ISO images, and other required files may prevent a safe unmount.
Power off, migrate, unregister, or relocate workloads and files according to your operational plan.

Not in a datastore cluster
A datastore cluster groups datastores for common management.
Remove the datastore from the cluster or resolve the cluster dependency before retrying.

Storage I/O Control disabled
Storage I/O Control manages storage resource allocation and latency behavior.
Disable it for the datastore before unmounting.

Not used for HA heartbeat
A vSphere HA heartbeat datastore supports host and virtual-machine state monitoring.
Choose another heartbeat location or remove this datastore from heartbeat use before proceeding.

No Storage DRS dependency
Storage DRS manages placement and balancing within a datastore cluster.
Resolve its association and any related automation before unmounting.

Affected hosts identified
A shared datastore may be mounted on multiple hosts.
Verify the host selection against the change plan; unmount additional hosts only when required.

Also check backup, replication, storage-management, host, and cluster workflows. Validate that backups are current and that all planned migrations have completed. A datastore that appears empty may still be required by an automation job, template, ISO repository, snapshot chain, or management feature.

For related background, review browsing VMFS datastores, datastore heartbeating, and deleting a VMFS datastore before choosing a destructive operation.

Unmount a VMFS datastore in the vSphere Client

  1. Sign in to the vSphere Client through vCenter Server with an account that has permission to manage the host and datastore.
  2. In the inventory, select the target ESXi host.
  3. Open the host's Related Objects view.
  4. Open Datastores to display datastores associated with the host.
  5. Select the VMFS datastore that you have verified is ready for unmounting.
  6. Use the datastore unmount action from the available vCenter actions or context menu. The exact label or menu location can vary by vSphere version.
  7. If the datastore is shared, select the individual ESXi hosts from which access should be removed. Do not select unaffected hosts merely because they can see the datastore.
  8. Read the warnings and confirmation prompt. Confirm that workloads, HA heartbeat use, Storage I/O Control, Storage DRS, datastore-cluster membership, and other dependencies have been resolved.
  9. Complete the operation and wait for the task to report success.

The navigation sequence is: select the ESXi host, open Related Objects, open Datastores, select the VMFS datastore, and choose the datastore unmount action from the available vCenter actions.

Post-unmount validation

  1. On each selected host, verify that the datastore no longer appears as mounted or accessible.
  2. For a shared datastore, check unaffected hosts and confirm that they can still see and use it when the maintenance plan intentionally removed access only from selected hosts.
  3. Confirm that the underlying storage contents remain intact through an approved storage or datastore inspection process. Do not infer data deletion from the datastore disappearing from one host.
  4. Record the successful host list and task result in the change record.

If access must be restored, restore storage presentation when necessary and mount the existing VMFS datastore again. Do not create a new datastore or format the backing device.

Practical examples

Removing a shared datastore from one host only

A shared VMFS datastore is accessible to three hosts, and one host is being retired. The administrator selects only the retiring host in the unmount workflow. The datastore remains mounted on the other two production hosts.

Preparing for storage-array retirement

After migrating all virtual machines and files to replacement storage, the administrator verifies backups, HA heartbeat settings, Storage I/O Control, Storage DRS, and datastore-cluster membership. The administrator then unmounts the datastore from every relevant ESXi host. Only after successful validation does the storage team proceed with device removal.

Resolving an unavailable unmount action

If unmounting cannot be completed because the datastore belongs to a datastore cluster, the administrator first removes it from that cluster and resolves associated Storage DRS usage. The administrator then checks the remaining dependencies and retries the operation.

Troubleshooting

The datastore cannot be unmounted

Likely causes include virtual machines or required files still using the datastore, datastore-cluster membership, enabled Storage I/O Control, HA heartbeat use, Storage DRS, or another active management dependency.

Identify the dependency, migrate or unregister workloads as appropriate, adjust the relevant datastore or cluster settings, and retry. Do not force the operation by detaching the device or removing the LUN first.

The datastore remains available on another ESXi host

This is expected when the datastore is shared and the unmount workflow targeted only selected hosts. Confirm the selected host list. Repeat the operation on additional hosts only if the maintenance plan requires removing access everywhere.

The administrator expected the data to disappear

Unmounting was confused with deletion or storage-device removal. The VMFS files remain on the underlying volume. Use a separately approved datastore cleanup, deletion, or device decommissioning process only when data removal is intended.

A host needs the datastore again

Restore storage presentation if it was removed, then mount the existing VMFS datastore. Never format the backing device to recover access to a datastore that was merely unmounted.

Operational cautions

  • Validate backups and migration completion before removing host access.
  • Make sure no host, cluster, backup, replication, or storage-management workflow still depends on the datastore.
  • For shared storage, distinguish the hosts that should lose access from those that must remain operational.
  • Unmount the datastore from every relevant host before detaching its storage device or removing its LUN presentation.
  • Keep the backing device or LUN identity in the change record so the correct storage is maintained or decommissioned.

Unmounting is a controlled access change, not a data-destruction operation. Treat later detachment, LUN removal, deletion, and formatting as separate changes requiring their own verification.