Failover Cluster-enabled VMs fail to power on due to VMDK format incompatibility

Products

• VMware vCenter Server
• VMware vSphere ESXi 8.0
• VMware vSphere ESXi

Issue/Introduction

Failover Cluster-enabled virtual machines fail to power on when using shared disks in multi-writer mode.

Symptoms

When attempting to power on the VM, the following error is displayed:

• Powering on the virtual machine fails:
  Status: VMware ESX cannot open the virtual disk "/vmfs/volumes/datastore/VM1/VM1_2.vmdk" for clustering. Verify that the virtual disk was created using the thick option. 

  Failed to start the virtual machine. Module Disk power on failed. Cannot open the disk '/vmfs/volumes/datastore/VM1/VM1_2.vmdk' or one of the snapshot disks it depends on. Thin/TBZ/Sparse disks cannot be opened in multiwriter mode File system specific implementation of OpenFile[file] failed

Additional symptoms:

• Failover Cluster-enabled VMs with shared VMDKs fail to power on.
• Shared disk type is reported as thin or lazy-zeroed thick instead of eager-zeroed thick.
• Failover clustering setup is unable to access the clustered volume.

Environment

VMware vSphere ESXi 7.0

VMware vSphere ESXi 8.0
VMware vCenter 7.0
VMware vCenter 8.0

Cause

The clustered VMDK was not provisioned as eager-zeroed thick, which is a strict requirement for multi-writer mode in vSphere.

• Failover cluster-enabled VMs with shared disks require eager-zeroed thick VMDKs.
• Thin, lazy-zeroed thick, or sparse formats are unsupported in multi-writer mode.

Resolution

Notes / Prerequisites

• The customer should take a full VM backup before conversion.
• Alternatively, a disk clone can be created with: (Reformat the VMDK using the vmkfstools -i command)
  vmkfstools -i source_VM_disk.vmdk target_VM_disk.vmdk -d eagerzeroedthick
  This creates a new, eager-zeroed disk. The existing disk can then be unmounted, and the newly created disk attached to the VM. 
  
  
• This KB is only applicable when customers do not have additional free space on the datastore or storage. In such cases, based on customer confirmation, an in-place disk conversion can be performed.

 

Steps

1. Check Disk Type
   SSH into the ESXi host and run:
   
   # vmkfstools -q /vmfs/volumes/<datastore>/VM1/VM1_2.vmdk
• Expected: eagerzeroedthick
  If reported as zeroedthick, proceed with conversion.
2. Power Off the VMs
• Ensure the VM is powered off.
• Confirm no snapshots exist.
3. Convert Disk to Eager Zeroed Thick (In-place)
   # vmkfstools -k /vmfs/volumes/<datastore>/VM1/VM1_2.vmdk
   
   What the command does:
• -k = eagerzero option.
• Converts an existing lazy-zeroed thick disk into eager-zeroed thick in-place.
• VMware sequentially writes zeroes to all unused blocks.
  
  Effects:
1. Disk type changes:
• Before → zeroedthick (lazy)
• After → eagerzeroedthick (verify with vmkfstools -q)
2. No data loss — existing guest OS data remains intact.
3. VM downtime required during conversion.
4. Duration depends on datastore throughput (writes across full disk size).
5. Storage impact: heavy sequential write I/O (other VMs on the datastore may experience temporary performance degradation).
   
   What it will NOT do:
• It will not shrink or expand the disk.
• It will not affect snapshots (conversion fails if snapshots exist — remove snapshots first).
• It will not work on thin disks (only valid for lazy-zeroed thick).
   
4. Detach and Reattach Disk
• End result: Your VM1_2.vmdk becomes eager-zeroed thick
• Remove the clustered VMDK from the VM (remove from VM only).
• Reattach it as an existing virtual disk.
   
5. Power On VMs
• Power on the Failover Cluster-enabled VMs.
• Validate disk access in the cluster.
  
  

• In-place conversion with vmkfstools -k changes the disk format from lazy-zeroed thick → eager-zeroed thick without data loss.
• This ensures compatibility with VMware multi-writer clustering while preserving existing data.
• After reattaching and powering on, the Failover Cluster-enabled VMs were able to access shared disks successfully and cluster health was validated.

Additional Information

• In-place conversion with vmkfstools -k changes the disk format from lazy-zeroed thick → eager-zeroed thick without data loss.
• This ensures compatibility with VMware multi-writer clustering while preserving existing data.
• After reattaching and powering on, the Failover Cluster-enabled VMs were able to access shared disks successfully and cluster health was validated.

Best Practices

• Always provision eager-zeroed thick disks upfront for any clustered or multi-writer workloads (such as Windows Failover Clustering, MSCS, SAP, etc.).
• Avoid using thin or lazy-zeroed thick disks for shared storage in cluster scenarios.
• Remove snapshots before enabling or modifying multi-writer disks.
• Ensure adequate datastore capacity and performance to handle eager-zero operations.
• Test changes in a staging or non-production environment before applying to production.
• Document disk types during VM build for clustered VMs to avoid future compatibility issues.