VMDK Timestamps Static During Guest OS Writes in Clustered Environments

Products

• VMware vSphere ESXi

Issue/Introduction

When virtual machines utilize shared storage architectures with Multi-writer locking or clustered filesystems (such as Oracle RAC or Microsoft WSFC with clustered VMDK), the modification timestamps (mtime) of the .vmdk files in the datastore directory do not update during active Guest OS disk writes. This is expected architectural behavior and should not be misinterpreted as a storage failure or hang.

Environment

• VMware vSphere ESXi 7.x / 8.x / 9.x
• VMware Cloud Foundation (VCF)
• VMware vSphere Foundation (VVF)
• VMFS6

Cause

Multi-writer locking disables exclusive host-level metadata ownership for the virtual disk to prevent contention between nodes. Because no single host maintains exclusive control of the file's management plane during concurrent access, the directory-level metadata attributes—including the mtime—are not updated dynamically on the datastore 

Resolution

This is expected architectural behavior for clustered and multi-writer configurations. To safely verify storage locking status or force a metadata update, follow these steps:

1. Inspect Lock State: Log in via SSH to the ESXi host registering the active VM and run the vmfsfilelockinfo utility:

   vmfsfilelockinfo -p /vmfs/volumes/####/####/####.vmdk

2. Force Timestamp Update: Gracefully power off or migrate all virtual machine nodes sharing the disk. This forces the hosts to close file handles and commit final metadata states to the directory structure.
3. Review Logs: For detailed troubleshooting, collect a log bundle from the host: Collecting diagnostic information for VMware ESX/ESXi using vm-support (313542).

Additional Information

• Enabling or disabling simultaneous write protection provided by VMFS using the multi-writer flag (313527)
• Note: For WSFC environments, the "Clustered VMDK" feature must be used instead of the multi-writer flag to avoid metadata inconsistencies