Inside the lab follow MDrepairs on YouTube, TikTok & Instagram - visit our YouTube channel ↗

Virtual Machine Recovery: Recovering Lost VMDK and VHD Files

Our founder and lead technician, Joseph Montanti, handles cases personally.

Virtual machine recovery for VMDK files starts with one fact: the data isn't gone just because the VM won't boot. A corrupted VMDK descriptor, a deleted VHD snapshot chain, a failed ESXi datastore. The virtual disk files themselves are just large container files sitting on physical storage. If the underlying drive (or RAID array, or SAN LUN) still has intact sectors, we can extract and rebuild those virtual disk images. That's true whether you're running VMware, Hyper-V, VirtualBox, or Proxmox.

We're MDrepairs, a professional data recovery lab based in New Jersey. We don't resell IT services. We don't run scripts and call it a day. We do hands-on data recovery at the hardware level, including full server recovery and SQL database recovery. When a virtual machine goes down and you can't get it back, that's exactly where we step in.

This guide covers everything: how virtual disk formats work, what causes VMDK and VHD corruption, the DIY fixes you should try first, and what happens when those don't work. If your VM holds production data and you can't afford to lose it, keep reading.

On this page

What Are VMDK and VHD Files?

Before we get into recovery, you need to understand what these files actually are. A VMDK (Virtual Machine Disk) or VHD (Virtual Hard Disk) is basically a container. It's a single large file on your physical storage that acts as an entire hard drive for a virtual machine. Everything inside that VM, the operating system, applications, databases, user files, all lives inside this one container file.

Think of it like a ZIP file, except instead of compressing random documents, it contains a full disk image. The VM hypervisor (VMware, Hyper-V, whatever you're using) reads and writes to this file as if it were a real physical hard drive.

VMDK (VMware)

VMware's format. Used by ESXi, vSphere, Workstation, and Fusion. A single VM might have multiple VMDK files. There's the descriptor file (a small text file with metadata) and the flat file (the actual data, sometimes called -flat.vmdk). If the descriptor gets corrupted or deleted but the flat file is intact, your data is almost certainly recoverable. That's actually one of the easier cases we see.

VMDKs can be "thin provisioned" (they grow as the VM writes data) or "thick provisioned" (pre-allocated to full size). Thin provisioned files are more fragile during a host crash because the file system has to track growing file boundaries dynamically. One bad shutdown and that tracking can go sideways.

VHD and VHDX (Microsoft Hyper-V)

Microsoft's format. VHD is the legacy format (max 2TB), and VHDX is the newer one (supports up to 64TB, better resilience to corruption). Hyper-V uses these, and so does Azure for cloud-based VMs. VHDX files include a built-in log that tracks in-flight updates, which means they're actually more resistant to corruption from power failures than older VHD files.

The log helps, but it can't save you from a physically failing drive or a RAID controller that scrambles the data layout.

Other Formats

VirtualBox uses VDI files. Proxmox and KVM use QCOW2. Citrix XenServer uses VHD internally. The recovery principles are the same across all of them. The container format is different, but the underlying concept is identical: a big file on physical storage that we need to extract and repair.

Format Hypervisor Max Size
VMDK VMware ESXi, Workstation, Fusion 62TB (ESXi 8.0)
VHD Hyper-V (legacy), VirtualBox 2TB
VHDX Hyper-V (current) 64TB
QCOW2 KVM, Proxmox, QEMU Effectively unlimited
VDI VirtualBox 2TB (VHD compat) or larger

Common Causes of VM File Loss and Corruption

Here are the scenarios that bring people to us for virtual machine recovery VMDK work.

1. RAID Array Failures

This is the big one. Most production VMs sit on RAID arrays (RAID 5, RAID 6, RAID 10). A RAID can tolerate one or two drive failures. But when a third drive fails, or when the RAID controller itself dies, the entire array goes offline. Every VM on that array goes with it.

The worst part? People often don't realize a drive has already failed in the array. The RAID was running degraded for weeks. Then a second drive fails and the whole thing collapses. No monitoring, no alerts, just a call one morning saying "everything is down."

Do NOT attempt a RAID rebuild with a failed controller. If your RAID controller died, do not swap in a different controller model and try to import the array. Different controllers use different stripe configurations. You'll make the data unrecoverable. Pull the drives and send them to a lab.

2. Corrupted VMDK Descriptors

The VMDK descriptor file is tiny. It's just a text file that tells VMware how to read the corresponding flat file. But if it gets corrupted, deleted, or mismatched, VMware can't mount the virtual disk. The VM won't power on. You'll see errors like "Cannot open the disk" or "The parent virtual disk has been modified since the child was created."

Good news here: if the flat file (-flat.vmdk) is intact, your data is fine. You just need to rebuild the descriptor. That's something a recovery tech (or a very careful admin) can do manually.

3. Snapshot Chain Corruption

Snapshots are useful, but they create chains of delta files. A VMware VM with three snapshots might have files like vm-000001.vmdk, vm-000002.vmdk, vm-000003.vmdk in addition to the base disk. Each delta file depends on the one before it. If any file in that chain gets corrupted, you can't commit or revert the snapshots, and the VM becomes stuck.

One corruption anywhere in that chain and everything grinds to a halt. If you're running production VMs, keep snapshots short-lived. Days, not months.

4. Accidental Deletion

Someone deletes the wrong VM from the datastore. Someone formats the wrong LUN. Someone runs rm -rf on the wrong directory. It happens more often than anyone wants to admit. In an ESXi SSH session, one wrong path and a 500GB VMDK is gone.

On traditional datastores (VMFS on spinning disks), deleted files can often be recovered because VMFS doesn't immediately overwrite freed blocks. On SSD-backed datastores or NVMe storage pools with TRIM enabled, the window is much shorter.

5. Power Failures and Host Crashes

A hard power loss while a VM is writing data can corrupt the virtual disk file. This is especially common with thin-provisioned VMDKs because the hypervisor was actively expanding the file when power cut. The file system metadata might say the VMDK is 200GB, but the actual allocated blocks on disk might only account for 180GB. That inconsistency makes the VM unable to start.

6. Storage Controller and Firmware Issues

Bad firmware updates on RAID controllers, HBA cards, or storage array controllers can corrupt data at the block level. The data is still on the physical disks, but the metadata that tells the controller how to assemble the blocks is wrong.

VM Layer vs Host Layer: Find Out Where the Failure Actually Lives

Every virtual machine sits on a stack of at least four layers, and a "lost VM" is really a failure at exactly one of them. Working on the wrong layer is the fastest way to turn a clean recovery into a lost cause, so this diagnosis comes before any tool touches anything.

From top to bottom: the guest file system (NTFS or ext4 inside the VM), the virtual disk file (the VMDK or VHDX itself), the datastore file system that holds those files (VMFS on ESXi, NTFS or ReFS on Hyper-V, or an NFS/iSCSI share on a NAS), and the physical storage underneath (a RAID array or individual disks). The symptoms tell you the layer: a VM that boots but shows corrupted data is a guest-level problem; a VM that will not start because its disk file is missing or unreadable is a virtual-disk or datastore problem; a whole datastore that vanished with every VM on it is the storage layer.

The one rule that protects every layer: work on copies. Copy the VMDK or VHDX off the datastore before repairing anything inside it. Running chkdsk in a guest whose virtual disk is corrupt, or letting a hypervisor "repair" a datastore that is really a failing RAID, writes fixes for one layer straight through the damage in another.

The layer also decides who can help. Guest and single-file problems are often solvable in place. But if the datastore lives on a failed array, the physical layer must be recovered first: drives imaged, the array reassembled virtually, the datastore extracted, and only then the VMDKs repaired. That order is fixed, and it is why we handle these cases bottom-up, as covered in our RAID 5 recovery guide and, for NAS-hosted datastores, the Synology NAS recovery guide. Tell whoever works on your case which hypervisor and which storage sat underneath: it changes the entire recovery sequence.

VMDK Recovery: VMware Virtual Disks

Here's how it breaks down by scenario.

Recovering a Corrupted VMDK Descriptor

If you've got the -flat.vmdk file but the descriptor is missing or corrupt, you can rebuild it manually. The descriptor is a plain text file that specifies the virtual hardware version, the extent (how big the disk is), and the parent file (for snapshots). Here's the general structure:

1 Check for the Flat File

SSH into your ESXi host and navigate to the datastore. Look for files ending in -flat.vmdk. If this file exists and is the expected size, your data is probably intact. Run: ls -lh /vmfs/volumes/datastore1/vmname/

2 Determine Disk Geometry

You need the exact size in sectors. Get the file size in bytes and divide by 512. For example, a 100GB flat file is 209715200 sectors. Run: stat -c %s vmname-flat.vmdk and divide by 512.

3 Create a New Descriptor

Create a new text file named vmname.vmdk with the correct descriptor format. The critical fields are createType, RW [sectors] VMFS "vmname-flat.vmdk", and the ddb.virtualHWVersion. Match the hardware version to your ESXi version (e.g., "20" for ESXi 8.0).

4 Register and Test

Add the VM back to inventory. Try powering it on. If the disk geometry matches, the VM should boot normally with all data intact.

Don't guess the sector count. If the descriptor specifies the wrong number of sectors, VMware will either refuse to open the disk or (worse) present corrupted data to the VM. Get the exact byte count of the flat file and do the math.

Recovering from Snapshot Chains

When a snapshot chain is broken, the safest approach is to clone the data out rather than trying to commit the snapshots. VMware's vmkfstools utility can clone a virtual disk to a new standalone VMDK, bypassing the broken snapshot chain entirely.

If vmkfstools -i fails because of the corrupted snapshot, you may need to manually edit the descriptor files to remove the parentFileNameHint references and consolidate the chain manually. This is tedious, error-prone work. One wrong edit and you've made things worse. If you're not confident with VMFS internals, this is where calling a professional makes sense.

Recovering Deleted VMDKs from VMFS

VMFS (VMware's file system) doesn't immediately zero out deleted file blocks. If a VMDK was recently deleted from a VMFS datastore, specialized tools can scan the VMFS metadata to find orphaned file entries. We use commercial forensic tools that understand VMFS structure at the block level. Open-source alternatives exist (vmfs-tools on GitHub), but they're limited compared to what we run in the lab.

Time is critical here. If the datastore is still active and VMs are still writing to it, those freed blocks will eventually get reused. The sooner you stop writing to that datastore, the better your chances.

VHD/VHDX Recovery: Hyper-V Virtual Disks

The recovery process is different from VMware, but the principles are the same.

VHD vs. VHDX Resilience

If you're running VHD (the old format), you're more vulnerable to corruption. VHD doesn't have a journal or log. A power failure mid-write can leave the internal metadata in an inconsistent state, and there's no built-in mechanism to repair it.

VHDX improved on this significantly. It has a built-in log region that records pending metadata changes before committing them. After a crash, Hyper-V can replay the log to restore consistency. That said, the log only protects VHDX-level metadata. It doesn't protect the file system inside the virtual disk (NTFS, ReFS, etc.) from corruption. So even with VHDX, you can end up with a mountable virtual disk that has a corrupted file system inside it.

Recovering Corrupt VHD/VHDX Files

Windows has a built-in tool. Open Hyper-V Manager, go to "Edit Disk," and select "Repair." This can fix some metadata inconsistencies. For VHDX files, it replays the log. For VHD files, it tries to rebuild the BAT (Block Allocation Table).

If the built-in repair doesn't work, you can try mounting the VHD/VHDX directly in Windows using Disk Management or PowerShell:

Mount with PowerShell

Run: Mount-VHD -Path "C:VMsdisk.vhdx" -ReadOnly

Using -ReadOnly is critical. You don't want Windows trying to "fix" the file system and making things worse. If it mounts, you can copy data out. If it doesn't mount, the corruption is at the VHD/VHDX container level, not just the file system inside.

Differencing Disks and Checkpoints

Hyper-V checkpoints (their word for snapshots) create differencing disks: AVHD or AVHDX files. These are delta files, just like VMware's snapshot deltas. Same problem applies. If a differencing disk in the chain gets corrupted, you can't merge the checkpoint and you can't revert to it.

The fix is similar to VMware: try to merge manually using Merge-VHD in PowerShell, or clone the data out at the block level. If the parent-child relationship in the chain is broken, you'll need to manually reconnect the chain by editing the parent locator entries in the AVHDX header. That's hex editor territory. Not something most admins should attempt without a backup plan.

RAID Array Failures and VM Recovery

Let's talk about the elephant in the room. Most VM data loss isn't really a VM problem. It's a storage problem. The VM files are fine conceptually. The physical storage they're sitting on has failed.

Why RAID Failures Kill VMs

A typical small business server runs 4-8 drives in RAID 5 or RAID 6. All the VMDKs or VHDs live on that array. When the array fails, every VM goes offline simultaneously. You're not recovering one VM. You're recovering the entire RAID first, and then extracting the VM files from the reconstructed array.

This is a two-phase process:

  1. Phase 1: RAID reconstruction. We image every drive in the array individually, then use the parity data and stripe configuration to reassemble the logical volume. If drives have bad sectors (they usually do, that's why the RAID failed), we work around them using our imaging tools.
  2. Phase 2: VM extraction. Once we have a coherent image of the array, we mount the file system (VMFS, NTFS, ReFS) and pull out the VM files. Then we verify each VM boots and the data inside is intact.
Never attempt to rebuild a degraded RAID by replacing the failed drive if multiple drives have failed. A rebuild puts massive read stress on the remaining drives. If another drive fails during rebuild (common with aging drives), the array is toast. Image the drives first, rebuild from images.

RAID Controller Failures

Sometimes the drives are perfectly fine but the RAID controller dies. This is tricky because different controller models (even different firmware versions of the same model) use different stripe sizes, parity rotations, and metadata layouts. You can't just plug the drives into a different controller and expect it to work.

We keep a library of RAID controllers in our lab specifically for these cases. If your Dell PERC H730 died, we can test with the same model. If we don't have the exact match, we reconstruct the RAID parameters manually by analyzing the data patterns on the drives. It's tedious work, but it gets results.

SAN and NAS Failures

Enterprise environments often run VMs on SAN storage (iSCSI, Fibre Channel) or NAS devices (Synology, QNAP, TrueNAS). When these fail, the recovery process depends on the underlying file system. Synology uses Btrfs or ext4. QNAP uses ext4 or ZFS. TrueNAS uses ZFS.

ZFS is interesting from a recovery standpoint. It has built-in checksumming and self-healing capabilities. If a ZFS pool degrades but doesn't fully fail, it can often heal itself when you replace the bad drive. But if the pool is completely lost (too many drives failed, or the pool metadata is corrupted), ZFS recovery is extremely specialized. The pool metadata structure is complex, and there aren't many consumer tools that handle it well.

DIY Recovery Methods You Should Try First

Not every VM problem needs a professional lab. Here are the things you should try before calling us. Be honest with yourself about your skill level though. If you're not comfortable with command-line tools and disk internals, skip to the professional recovery section.

Check Your Backups First

Sounds obvious, right? You'd be amazed how many people panic and start attempting recovery before checking if they have a backup. Check Veeam, Nakivo, Altaro, Windows Server Backup, whatever you use. Check offsite copies. Check cloud replicas. If you have a backup from before the corruption, restore from that. It's faster, safer, and cheaper than any recovery method.

VMware: Rebuild the Descriptor

If the flat file is there but the descriptor is missing, follow the steps in the VMDK recovery section above. VMware has a KB article (KB1002511) that walks through it.

VMware: Clone Past the Corruption

If a VMDK is partially corrupted, try cloning it with vmkfstools:

vmkfstools -i /vmfs/volumes/datastore1/vm/vm.vmdk /vmfs/volumes/datastore2/vm-recovered/vm.vmdk -d thin

This reads the source VMDK and writes it to a new location. If the corruption is in non-critical areas, the clone may succeed and produce a working copy.

Hyper-V: Inspect and Repair

Use the Hyper-V "Inspect Disk" and "Edit Disk > Repair" tools. For VHDX files, this replays the journal log. For checkpoint issues, try merging checkpoints through PowerShell:

Merge-VHD -Path "C:VMsvm-checkpoint.avhdx" -DestinationPath "C:VMsvm.vhdx"

If this fails with a parent-child mismatch error, the chain is broken and you'll need manual repair or professional help.

Mount and Copy

Regardless of hypervisor, you can try mounting the virtual disk file directly on another machine and copying data out. On Linux, use qemu-nbd to mount VMDKs, VHDs, and QCOW2 files. On Windows, use Disk Management for VHDs or third-party tools like 7-Zip (yes, it can open VMDK files) for simple file extraction.

Always work on a copy. Before attempting any repair or recovery on a corrupted VM file, make a byte-for-byte copy first. Use cp, robocopy, or dd. If you make the corruption worse, you want an untouched copy to fall back on.

Free and Open Source Tools

  • TestDisk / PhotoRec: Free, open-source. Can scan raw disk images (including mounted VMDKs) for recoverable files. Won't fix the VM, but can pull individual files out.
  • R-Studio: Not free, but very capable. Understands VMFS, VMDK, VHD, and most file systems. Can scan a damaged virtual disk and extract files with folder structure intact.
  • vmfs-tools (Linux): Open-source utilities for reading VMFS datastores on Linux. Useful if your ESXi host is dead but you have the raw drives.
  • libguestfs: Linux toolkit for accessing and modifying VM disk images. Can mount VMDKs and VHDs directly without the hypervisor.

When You Need Professional Virtual Machine Recovery

DIY methods work for logical corruption. Missing descriptors, broken snapshot chains, accidentally deleted files on a healthy datastore. But there are situations where you need a professional data recovery lab. No amount of software can fix hardware.

You Need Us When:

  • The physical drives have failed. Clicking, not spinning, not detected by the host. The VM files exist on drives that can't be read normally.
  • The RAID array has collapsed. Multiple drive failures, dead controller, or corrupted RAID metadata. Rebuilding a RAID from damaged drives requires specialized imaging equipment.
  • The datastore is formatted or overwritten. Someone formatted the LUN. A new OS installation wiped the partition table. The VM files were overwritten by a different installation.
  • Software recovery has already failed. You've tried the DIY steps. The tools can't read the data. The corruption goes deeper than what software can fix.
  • The data is business-critical. Production databases, financial records, years of company data. The stakes are too high for trial-and-error.

What Our Lab Does Differently

We're not running the same software tools you'd find on Google. Our process for virtual machine recovery VMDK cases involves:

  1. Hardware-level drive imaging. We use dedicated imaging tools (PC-3000, DeepSpar) that can read drives with bad sectors, weak heads, and firmware issues. These tools read at the physical level, bypassing the drive's own error handling. A drive that won't mount on your server might yield 99.9% of its data on our equipment.
  2. RAID parameter detection. For RAID arrays, we analyze the raw data patterns to determine stripe size, parity rotation, drive order, and offset. We don't guess. We verify mathematically before assembling the array.
  3. File system reconstruction. Once we have the raw data, we reconstruct VMFS, NTFS, ReFS, ZFS, or whatever file system the VMs were stored on. We extract the VMDK/VHD files intact.
  4. VM verification. We don't just hand you a pile of files. We mount each recovered VM, boot it in a test environment, and verify the data inside is accessible and intact. You get working VMs back, not guesswork.
Our track record: We maintain a better than 90% success rate on virtual machine recovery cases where the physical media arrives in our lab without prior recovery attempts. That drops significantly when someone has already tried rebuilding the RAID or running consumer recovery software on failing drives. Don't make things harder for us. Stop using the drives and send them in.

Snapshot and Backup File Recovery

Sometimes the VM itself is gone, but snapshots or backup files survive. These can be your lifeline.

VMware Snapshots

VMware snapshot delta files (-000001.vmdk, etc.) contain only the changes made since the snapshot was taken. They're useless without the base disk. But if you have the base disk and at least some of the delta chain, we can merge whatever exists and recover the VM up to the last intact snapshot.

Veeam Backup Files

Veeam stores backups as VBK (full backup) and VIB/VRB (incremental) files. These are essentially compressed and deduplicated copies of the VM's virtual disks. If your primary storage is gone but Veeam backups exist on a different drive or NAS, we can extract the VM data from Veeam's proprietary format.

Windows Server Backup

Windows Server Backup stores Hyper-V backups as VHD files inside a WindowsImageBackup folder. These are standard VHDs that can be mounted directly. If you can find these files on a backup drive, you might not need us at all. Just mount the VHD and copy your data out.

Preventing VM Data Loss

Recovery works, but prevention is better. Here's what we recommend based on the patterns we see in failed environments.

Backup Strategy: 3-2-1 Rule

Three copies of your data, on two different types of media, with one copy offsite. For VMs, this means:

  • Primary: The live VM on your production storage.
  • Local backup: Veeam, Nakivo, or similar backing up to a separate NAS or DAS.
  • Offsite/cloud: Replicate to AWS S3, Azure Blob, Backblaze B2, or ship drives to a second location.

If this sounds expensive, think about what it costs to lose every VM in your environment. We charge for recovery. Backup software is cheaper.

Monitor Your RAID

Set up SNMP alerts or email notifications on your RAID controller. Every controller supports this. You should know about a failed drive within minutes, not weeks. A RAID 5 running degraded for a month is a ticking bomb.

Limit Snapshots

Snapshots aren't backups. They're short-term rollback points for patches and updates. Take a snapshot before a Windows update, verify everything works, then delete the snapshot within a day or two. Long-running snapshot chains cause performance issues and increase the blast radius when something goes wrong.

VMware recommends a maximum of 32 snapshots per VM. We recommend no more than 3. And never leave them running for more than 72 hours unless you have a very specific reason.

Test Your Restores

A backup that's never been tested is just a theory. Every quarter, pick a random VM and restore it to an isolated network. Boot it. Verify the data. If you can't restore it in a test, you won't be able to restore it in an emergency.

Quick sanity check: Can you name, right now, the exact location of your most recent VM backup? Can you confirm it completed successfully? If the answer to either question is "I'm not sure," fix that before anything else.

Use UPS Protection

Power failures cause VM corruption. Full stop. Every server hosting VMs should be on a UPS with automatic graceful shutdown configured. Include this in your disaster recovery plan. The UPS doesn't need to run the server for hours. It just needs to buy enough time (5-10 minutes) for the hypervisor to cleanly shut down all VMs before the battery runs out. A $300 UPS can save you $3,000+ in recovery costs.

Keep Firmware Updated (Carefully)

RAID controller and HBA firmware updates should be applied, but test them in a non-production environment first if possible. Read the release notes. If the update fixes a bug you don't have, consider waiting.

Frequently Asked Questions

Can you recover a deleted VMDK file from an ESXi datastore?

Yes, in most cases. When a VMDK is deleted from a VMFS datastore, the data blocks aren't immediately overwritten. We use forensic tools that scan VMFS at the block level to locate and reconstruct deleted VMDK files. The key factor is time: the sooner you stop using the datastore after deletion, the better the chances. If other VMs are still running on the same datastore and writing new data, those blocks may get reused. Power down non-essential VMs and contact us immediately. Taking the datastore offline avoids further writes but does not guarantee recovery.

How long does virtual machine recovery take?

Our standard recovery estimate is 4–6 weeks; timing may vary by case. A full RAID reconstruction with multiple failed drives and terabytes of data can take longer. The diagnostic (where we evaluate the drives and tell you what's recoverable) depends on the queue and device condition. We'll give you a timeline and cost estimate for approval before recovery work starts.

Is virtual machine recovery VMDK data confidential?

Absolutely. Financial databases, medical records, legal documents. We can sign an NDA before you ship the drives. All recovered data is transferred to encrypted media, and we securely wipe our work copies after you confirm receipt. Our lab has physical security controls and restricted access.

Can you recover VMs from a failed Synology or QNAP NAS?

Yes. Synology, QNAP, TrueNAS, Drobo, and others. Most NAS devices use Linux-based RAID (mdadm) with ext4, Btrfs, or ZFS file systems. We pull the drives, image them individually, reconstruct the RAID, mount the file system, and extract the VM files. The process is the same whether the VMs were running on Synology's Virtual Machine Manager, QNAP's Virtualization Station, or were stored as backup files on the NAS.

What's the difference between recovering a VMDK and recovering a VHD?

The recovery principles are the same, but the file formats are different internally. VMDK files (VMware) use a descriptor/flat file structure, while VHD/VHDX files (Hyper-V) use a BAT (Block Allocation Table) structure. VHDX files are generally more resilient due to their built-in log, but both formats are recoverable from failed storage. The bigger factor in recovery success isn't the file format. It's the condition of the physical drives and the RAID configuration. A healthy VMDK on a dead RAID and a healthy VHD on a dead RAID require the same physical recovery work.

Do you recover VMs from SSDs?

We do, but SSD recovery is harder than HDD recovery in general. SSDs use TRIM, which actively erases deleted data blocks. If a VM was deleted from an SSD-backed datastore with TRIM enabled, the data blocks may have already been wiped by the drive's garbage collection process. That said, not all enterprise storage environments use TRIM (many SANs disable it), and enterprise SSDs sometimes retain data longer than consumer drives. We evaluate each case individually. Check out our data recovery services page for more on what we recover.

Can I ship drives to your lab for VM recovery?

Yes. We accept drives shipped from anywhere in the US. Pack each drive individually in anti-static bags with bubble wrap or foam padding. Don't use packing peanuts (they create static). Label each drive with its position in the RAID array if possible. Include a note describing what happened, what hypervisor you were running, and the RAID configuration (RAID 5, RAID 6, etc.). This information helps us start the recovery faster. We'll provide a prepaid return shipping label for the recovered data.

Corrupted Virtual Disk Recovery for VMDK, VHD, and VHDX Files

A corrupted virtual disk is not the same as a single damaged document. A VMDK, VHD, or VHDX file contains an entire disk structure inside a container: partition table, file system, operating system files, application data, databases, snapshots, and sometimes multiple linked extents. When that container is damaged, the virtual machine may refuse to boot, mount as RAW, show missing partitions, or report snapshot chain errors.

The first step is to preserve every file that belongs to the virtual machine. Do not create a new VM over the old folder, consolidate snapshots, delete delta files, or let the hypervisor attempt automatic repairs until a copy is made. VMware, Hyper-V, and VirtualBox can change metadata when a VM is opened, especially if snapshots are present. A small descriptor file may point to large flat or delta extents, so missing or renamed companion files can make the disk appear worse than it is.

Common causes include host drive failure, interrupted snapshot consolidation, datastore corruption, accidental deletion, thin-provisioned disk problems, backup software errors, and running out of space during writes. Recovery depends on the exact structure. For VMware, the descriptor, flat file, delta files, and snapshot order may need to be rebuilt. For Hyper-V, AVHDX checkpoints may need to be matched correctly to the parent VHDX. For VirtualBox, differencing disks can require parent media identifiers and chain repair before file recovery is attempted.

A lab process works from copies and disk images, then reconstructs the virtual disk chain before extracting guest files or restoring the VM. If the host storage is failing, the host drive or RAID should be imaged first because the virtual disk may only be one layer of the problem. MDrepairs handles corrupted virtual disks from local drives, NAS devices, RAID arrays, and external backup media. For VM-specific cases, use our virtual machine recovery service before opening or consolidating the VM again.

Can a corrupted virtual disk be repaired without losing files?

Sometimes, but the safest approach is to copy or image the VM files first and repair the copy. Direct repairs can change descriptors, snapshots, and allocation tables. If the corruption is logical, the virtual disk chain may be rebuilt. If the host drive is failing, the underlying storage must be stabilized before VM recovery.

What files are needed to recover a corrupted VMDK?

For VMware, keep the entire VM folder if possible. That can include the small descriptor VMDK, flat VMDK, delta VMDK files, VMX configuration, snapshot files, and logs. Do not rename or delete companion files. Even small metadata files can be important for rebuilding the correct disk geometry and snapshot order.

Get help with data recovery

Bring your device to our Lincroft lab during opening hours, or mail it in.

How to mail your device

The $100 diagnostic is non-refundable once diagnostics start, and donor drives or replacement parts you approve are non-refundable once ordered, even if recovery fails.

Published by MDrepairs

MDrepairs is a data recovery and electronics repair lab in Lincroft, New Jersey. Our founder and lead technician Joseph Montanti films real cases for our YouTube channel, showing tools, repair steps and recovery outcomes.

Keep reading

Data recovery services

Devices recovered at our lab

These rows come from BenchKey, the ticket system we built for our own shop. Customer names, contact details and descriptions are left out.

Portable hard drive circuit board with a USB socket
At our Lincroft lab, the USB socket is the drive's connection point to a computer.
4.6 · 345 Google reviews$100 diagnostic, credited toward a successful recovery
Start your mail-in recovery