A host-based backup job left my Hyper-V VM with a checkpoint chain that would not merge through the management interface. I shut down the VM, identified the AVHDX chain, and manually merged each differencing disk into its parent until only the original VHDX remained. However, Hyper-V still lists the old checkpoints, and the VM will not start because its configuration points to an AVHDX file that no longer exists. Should I reconnect the merged VHDX to the existing VM, remove the stale checkpoint metadata, or create a new VM?
3 Answers
Keep the existing VM and reconnect the merged VHDX rather than rebuilding it. That preserves the VM identifier, network configuration, and any backup jobs associated with it. With the VM powered off, use Get-VMHardDiskDrive to find the disk attachment, then use Set-VMHardDiskDrive with the same controller and location but point it at the merged VHDX. Once the disk reference is corrected, the VM should boot. After that, remove the stale checkpoints with Get-VMSnapshot and Remove-VMSnapshot. If the entries are recovery checkpoints and removal fails because their files are missing, a clean replacement VM using the same generation and firmware settings is the fallback. After recovery, run a new full backup because the manually merged chain should not be trusted for incremental backups.
The lingering checkpoints are probably related to the host-based backup workflow leaving a disk chain behind after an interrupted job. Check which checkpoint type the backup process creates. Production checkpoints use guest-level services and generally avoid leaving the same kind of standard AVHDX chain, while configuring ProductionOnly makes the operation fail instead of silently falling back. Once the disk is reconnected, check the checkpoint list with PowerShell as well as the management console; sometimes the console needs to be reopened before its display updates. Run a fresh backup and verify a restore afterward, and investigate why the original backup job abandoned the checkpoints.
These appear to be lingering checkpoints created by the host backup agent. I have not seen the same problem when the backup agent runs inside the guest, so I may switch to that approach after finishing the cleanup and testing a restore.
You may also be able to select the stale checkpoint in the management console and press Delete. Some backup-created checkpoints do not show the usual context-menu options, but the keyboard delete action can still trigger the cleanup and merge process.
I tested reconnecting the merged disk on an exported copy, and that part worked—the VM booted. Removing the recovery checkpoints still fails because Hyper-V cannot find the referenced files, so I am testing other cleanup options before creating a replacement VM.