VMware Datastore Missing or Will Not Mount
An ESXi host can lose a datastore even while it still detects the backing device. The failure may be in the RAID, SAN, NAS, LUN presentation, partition table, VMFS metadata, or storage path between the host and the underlying members.
Do not create a new datastore, resignature the only copy, format the device, or run repair against uncertain backing storage. Preserve the host and storage evidence first.
What this condition means
A missing datastore is a dependency-chain failure, not a single diagnosis. ESXi may have lost the device, detected a snapshot LUN, rejected damaged VMFS metadata, or received a changed logical volume from the controller or storage system.
The correct starting layer matters. VMFS work cannot compensate for an unstable RAID member, incomplete SAN presentation, or backing volume assembled from the wrong member state.
What to record
- ESXi and vCenter versions and exact datastore messages
- Datastore name, VMFS version, UUID, and last known capacity
- Backing local RAID, SAN, NAS, HBA, controller, or LUN details
- Affected hosts and whether the device is visible to each one
- Power, firmware, controller, path, rebuild, snapshot-LUN, mount, or resignature actions
What not to do
- Create or format a datastore on the affected device
- Force a resignature without identifying the presented volume
- Run VMFS repair against unstable backing storage
- Rescan and reboot repeatedly while device identity changes
- Write recovered files back to the affected datastore
How ADR evaluates it
ADR maps the path from physical members through RAID or shared storage to the LUN, partition, VMFS datastore, virtual disks, and priority VMs. Each layer is validated before work moves upward.
Stable logical cases may qualify for remote recovery. If the backing storage is unstable or the logical volume is uncertain, protected imaging and reconstruction come before VMFS or VMDK work.
What happens next
A datastore that will not mount can still contain recoverable virtual machines. Results depend on the backing-storage state, VMFS metadata, prior writes, and whether the correct device is being presented.