Australian cloud for business software — built and run in Sydney since 2002

NetLeverage
NetLeverage

VM will not power on: Cannot open the disk or one of the snapshot disks it depends on

Technical article
VMware

About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.

A power-on fails with Reason: Cannot open the disk disk_name or one of the snapshot disks it depends on. (Broadcom KB 313078). That line is a wrapper. The fault is in the line printed beside it, such as "Failed to lock the file" or "The parent virtual disk has been modified since the child was created". Read that line first, check the files without changing anything, and fix only the cause it names.

What the error means

Broadcom's KB 313078 shows the full message as An unexpected error was received from the ESXi host while powering on VM <VM_name>. followed by the Reason line. It lists seven values for the reason:

  • Failed to lock the file.
  • The parent virtual disk has been modified since the child was created.
  • The destination file system does not support large files.
  • Could not open/create change tracking file.
  • Cannot allocate memory.
  • The file specified is not a virtual disk.
  • Insufficient permission to access file.

"One of the snapshot disks it depends on" means the VM's disk is a chain: a base disk, plus a delta disk for each snapshot. If any link cannot be loaded, the whole disk cannot be loaded. The VM's vmware.log carries the same text. In Broadcom's examples it appears as [msg.disk.noBackEnd] Cannot open the disk ... or one of the snapshot disks it depends on. followed by [msg.disk.configureDiskError] Reason: ....

A missing file is not one of the seven Reason values. It appears in log lines, such as VMware ESX cannot find the virtual disk and The system cannot find the file specified (KB 367921 and KB 368913).

NetLeverage runs VMware vCenter, vSphere and ESXi.

Likely causes

Cause What you see How to confirm
Another process holds a lock Reason: Failed to lock the file. Often after a failed or interrupted backup vmfsfilelockinfo names the host holding the lock. A backup proxy VM may still have the disk attached
Snapshot chain mismatch Reason: The parent virtual disk has been modified since the child was created Content ID mismatch in vmware.log. A child's parentCID differs from its parent's CID
A file in the chain is gone Log shows The system cannot find the file specified A gap in the -00000N numbering in the VM folder
Leftover change tracking reference Reason: Could not open/create change tracking file DISKLIB-CTK lines in vmware.log
Descriptor file damaged or missing Reason: The file specified is not a virtual disk The .vmdk descriptor is corrupt or missing
NFS permissions Reason: Insufficient permission to access file, on an NFS datastore The export's read/write permissions for the host
VMFS heap exhausted Reason: Cannot allocate memory Host warnings about VMFS heap

Prove the cause before changing anything

All of these are read-only. Do not keep retrying the power-on. Broadcom's KB 345254 warns that when a VM is powered on, the CID of the disk it references is changed, and that further irreversible changes can follow.

1. Read the reason. In the vSphere Client task details, or in the VM's log (KB 322285 gives the path as /vmfs/volumes/datastore/VM-Name/vmware.log):

grep -n -E 'Cannot open the disk|Reason:|Content ID mismatch|change tracking|Locking conflict|cannot find' /vmfs/volumes/ds-prod01/web01/vmware.log

2. List the VM folder and look at the snapshot numbering:

ls -l /vmfs/volumes/ds-prod01/web01/

Each snapshot has a descriptor (web01-000001.vmdk) and a data file (web01-000001-delta.vmdk). Broadcom's KB 417298 says a jump in the numbering, for example from -000003 to -000005, shows a missing snapshot disk.

3. Change to the VM folder. Steps 4 to 6 use file names without a path, so run them from the VM's own folder. This command only changes the shell's current directory:

cd /vmfs/volumes/ds-prod01/web01

4. Read the descriptors, and only the descriptors. Files named -flat.vmdk or -delta.vmdk hold disk data and are not text. Name each descriptor in the command:

grep -E '^(CID|parentCID|createType|parentFileNameHint)' web01.vmdk web01-000001.vmdk web01-000002.vmdk

A good chain: the base disk has parentCID=ffffffff, and each child's parentCID equals the CID of the file its parentFileNameHint names. A bad chain: one child's parentCID differs from its parent's CID. KB 345254 shows that example.

5. Trace the chain. KB 414970 gives this command with the -000001.vmdk descriptor of its example VM, run in the VM folder, and says to look for "File not found" or "Inconsistent parent" messages. Use the top descriptor of your own chain, which is the one the VM's configuration names for the disk:

vmkfstools -q -v10 web01-000002.vmdk

6. Find the lock owner. KB 322285 shows this form for a locked disk:

vmfsfilelockinfo -p web01-flat.vmdk

Its output names the host's MAC address and the lock mode. The guide "A VMDK says it is locked and I need to find out who holds the lock" covers reading and clearing locks. Also check backup proxy VMs: KB 367921 and KB 396139 describe a proxy whose Edit Settings, Hard disks list still holds disks of other VMs.

Fix

Match the cause. Several of Broadcom's procedures edit or move VM files, and they are described below and not given as steps. They belong with your support contact or a change process.

  • Chain mismatch. KB 345254 describes correcting the CID or parentCID value in a descriptor after copying the files. It warns that "the guest operating system may be unable to boot successfully even after making corrections." If snapshot disks are missing, KB 417298 says "The virtual machine must be restored from a recent backup to ensure a consistent operating state."
  • Change tracking. KB 322285 moves the -ctk.vmdk files aside, sets ctkEnabled to false for each disk in the .vmx, and notes that disabling and enabling CBT "will result in full backup." KB 313078 refers to a separate KB for the "could not open/create change tracking files" error.
  • Not a virtual disk. KB 313078 says to create a new descriptor file and gives a separate KB for it.
  • NFS permissions. KB 313078 says to ensure the host has read/write permission on the NFS export, and that if the array has a No Root Squash option it should be enabled.
  • Cannot allocate memory. KB 334252 refers to a KB on VMFS heap warnings, and to a KB on committing snapshots when Snapshot Manager shows none.
  • Locked by a backup proxy. KB 367921 and KB 396139 detach the other VM's disks from the proxy. That is the one fix given as a step, in the next section.

Detach another VM's disks from a backup proxy

KB 396139 says: "Ensure no active backup is running while changes are being made."

First save the proxy's current disk list. In the vSphere Client, right-click the proxy VM, choose Edit Settings, and write down every hard disk entry with the full path of its VMDK file.

Risk: the next steps remove hard disks from the proxy VM. If "Delete files from datastore" is ticked, the VMDK file is deleted from the datastore, and for a disk that belongs to another VM that is that VM's data.

Follow KB 367921 and KB 396139: keep only the disks native to the proxy. For each hard disk that belongs to another VM, click the X icon to detach it. Check that "Delete files from datastore" is unticked. Click OK, then power on the VM that failed.

To undo it, only if the VM that owns the disk is powered off and your earlier ls -l of the owning VM's folder shows the file still there: edit the proxy's settings, click Add New Device on the Virtual Hardware tab, choose Existing Hard Disk, select that file from the list you saved, and click OK. Broadcom's vSphere 8.0 page for adding an existing disk gives these steps. If the owning VM is running, do not re-attach its disk to the proxy, because that recreates the lock this fix removes. In that case, or if the file has been moved or deleted, leave the disk detached and keep the saved list as the record of what the proxy held.

Example (illustrative): web01 is on datastore ds-prod01 and fails to power on with Reason: Failed to lock the file. vmware.log contains a Locking conflict line. ls -l shows a normal folder, and the descriptors show consistent CID and parentCID values, so the chain is sound. vmfsfilelockinfo -p web01-flat.vmdk names the host holding the lock. A backup job on proxy01 failed the night before. In Edit Settings for proxy01, Hard disk 2 points at [ds-prod01] web01/web01.vmdk. The administrator records the list, detaches that one disk with "Delete files from datastore" unticked, and powers on web01. It starts.

Confirm the fix

  1. The power-on succeeds and the VM's guest operating system starts.
  2. vmfsfilelockinfo -p web01-flat.vmdk no longer reports a lock held by another host.
  3. For a chain fault, vmkfstools -q -v10 on the top descriptor shows no "File not found" or "Inconsistent parent" messages.
  4. The task list shows no new Cannot open the disk error.

What to watch afterwards

  • Check backup proxy VMs after every failed backup. Compare their hard disk lists with the disks they own.
  • Do not resize, move or edit a disk that has snapshots. KB 345254 lists modifying disks with existing snapshots, and adding snapshotted disks to other VMs, among the scenarios to avoid.
  • After a Storage vMotion or a migration is interrupted, read the VM folder before powering on.
  • If the VM runs on a snapshot, read the guides "vCenter says "Virtual machine disks consolidation is needed": why, and scenario-specific recovery decisions to clear it" and "The -000001.vmdk and delta files in a VM folder: what each one is".

Risks and limits

  • These commands have not been run or lab-tested.
  • The Reason list in KB 313078 was updated 02-26-2025 and gives no version range. KB 313078 also lists an older hardware version and Windows Server Failover Clustering placement as causes, which are not covered here.
  • Descriptor edits, change tracking file moves and descriptor re-creation are described, not instructed. They change the files that hold the VM's disk, and some cannot be undone once the VM has been powered on.
  • A chain with missing snapshot disks cannot be repaired by editing IDs. Broadcom's KB 417298 says the VM must be restored from backup.

Sources