Fix it now
The job died in the shadow copy stage, so nothing was written. Event ID 521 is the one that carries the real error: “Backup started at <time> failed as Volume Shadow copy operation failed for backup volumes with following error code <code>”. Check the destination and the writers before anything else.
vssadmin list writers
vssadmin list providers
vssadmin list shadowstorage
fsutil dirty query C:
- Read the numeric code inside Event ID 521 first. It is the actual error; 0x807800C5 on its own tells you only that the snapshot stage failed.
- Check the destination, which is a real cause and not a distraction: rename or remove an existing WindowsImageBackup folder on the target, confirm the target is NTFS with ample free space, confirm no shadow copy exists on the target volume, and try a different target disk.
- Restart the service behind any writer whose state is not [1] Stable, then re-run
vssadmin list writersand confirm it came back clean. - If a third-party backup product left a shadow copy provider registered on a server that no longer uses it, uninstall the product properly rather than deleting files.
- Confirm no source volume is flagged dirty, and check free space on every source volume including the small hidden ones.
If the target is a network share, 0x8078015B is a different failure with a different fix – Windows Backup could not access the remote shared folder. Check the share is reachable and writable over SMB, not that it has room.
If a full job completes and its detail log is clean, you can stop here. The next section explains why one writer or one volume takes the whole job with it.
Why it happens
Windows Server Backup does not read your live files. It asks the Volume Shadow Copy Service for a point-in-time copy of every volume in the backup set and reads from that copy while the server carries on. Building it is a coordinated operation: VSS asks each registered writer to bring its data to a consistent state, holds writes briefly, creates the shadow copies, then releases everything.
That coordination is all-or-nothing across the snapshot set. If one writer fails, the whole set is abandoned rather than produce a copy that is consistent for some data and not for others – Microsoft’s own wording is that if any of the VSS writers encounter an error, the entire backup job will fail. The engine then reports that it could not prepare the backup image, and Event ID 521 records the underlying code.
Reading 521 properly is the difference between an hour and an afternoon. Its text carries a decimal number, and in Microsoft’s own worked example that number is 2155348129 – which is 0x807800A1. Convert it, and you are looking at the actual failure rather than at the generic one.
The destination is the problem after all
You have this one if The job fails at the same point every time, and nothing about the writers or the source volumes has changed.
- Rename or remove the existing WindowsImageBackup folder on the target, so the job creates a fresh one.
- Confirm the target volume is NTFS and has ample free space.
- Confirm there is no shadow copy on the target volume itself.
- Try a different target disk before spending time on writers – a failing external disk produces this reliably.
This check is cheap and it is often the answer. Do it before the deep VSS work rather than after it.
A writer is stuck in a failed state
You have this one if vssadmin list writers shows a writer whose state is not [1] Stable, and the same one fails every attempt.
- Identify the service that owns the writer and restart that service.
- Re-run
vssadmin list writersand confirm the writer is back to [1] Stable with no last error. - Re-run the backup immediately, while the writer is freshly reset.
- If it fails again straight away, read the owning application’s own log. VSS is reporting that application’s problem, not creating it.
Some writers only re-register at startup. If a service restart does not clear it and the server can be rebooted, do that before spending an afternoon.
The target is a network share the server cannot reach properly
You have this one if 0x8078015B rather than 0x807800C5, on a job that writes to a UNC path.
- Confirm the share is available and writable from the server itself, not just from your workstation.
- Confirm the file server answers over SMB and not merely over TCP: the Server service running and responsive, and disks that are not so slow they make SMB appear hung.
- Where the failure is latency, set
SessTimeoutto 1200 as a REG_DWORD underHKLM\SYSTEM\CurrentControlSet\Services\LanmanWorkstation\Parametersand restart the server. - Re-run the job and watch whether it now reaches the write stage.
A leftover provider is intercepting the snapshot
You have this one if vssadmin list providers shows a provider from a backup product that is no longer in use.
- List the providers and identify anything that is neither the Microsoft software provider nor your storage vendor’s supported hardware provider.
- Uninstall the product that registered it, properly, rather than deleting files.
- Reboot and confirm the provider list is what you expect.
- Re-run the backup.
Two backup products on one server is a reliable way to produce this. Only one should own the snapshot path.
Full reference
Splitting the failure by where it is reported
| What you find | Which layer to work on |
|---|---|
| A writer not in [1] Stable | The application that owns that writer |
A provider you do not recognise in vssadmin list providers |
The provider layer, usually a product that was removed badly |
| A source volume flagged dirty | File system damage blocking the snapshot |
| 0x8078015B in the log | The remote share used as the target, not the source volumes |
| Failures only when one particular volume is included | That volume, its shadow storage association or its free space |
| Failures only during business hours | Contention rather than configuration |
Shadow copy storage, and moving it safely
Every volume in the backup set needs somewhere to hold the original contents of blocks that change while the snapshot exists. vssadmin list shadowstorage /for=X: shows the association and its Maximum Shadow Copy Storage Space. Two operations change it and they are not the same:
| Command | What it does |
|---|---|
vssadmin resize shadowstorage /for=X: /on=Y: /maxsize=<size> |
Resizes the maximum size of an association that already exists. Both /for and /on are required, and they must name the pair that exists |
vssadmin add shadowstorage /for=X: /on=Y: /maxsize=<size> |
Creates an association so that volume X’s diff area lives on volume Y. Not listed on Microsoft’s current vssadmin reference, which documents four subcommands |
vssadmin list shadowstorage |
Shows the associations, with used, allocated and maximum sizes |
Microsoft’s own warning on resizing is that it may cause shadow copies to disappear. On a server where the existing restore points are the only copy of something, that is a data-loss step run blind. Check what depends on the current shadow copies first, and note that the documented maximum can be given as a size, as a percentage, or as UNBOUNDED.
The events and codes you will meet
| Code or event | What it is |
|---|---|
| Event ID 521 | Backup failed because the Volume Shadow copy operation failed for the backup volumes, with the real code in the text |
| 0x8078015B | Windows Backup could not access the remote shared folder used as the target |
| 0x807800C5 | No published meaning. Treat it as “the snapshot stage did not produce a usable copy” and work from 521 |
| Event ID 12293 | Microsoft’s Hyper-V backup guidance reads a 12293 in a guest, with 8229 on the host, as VSS writer errors |
| Event ID 517 | Referenced by Microsoft as a Windows Server Backup failure entry, with no published message text |
Where the evidence lives
- The Application log, for VSS events and for the writer’s own application errors in the same second.
- The Microsoft-Windows-Backup log, for 521 and its numeric code.
- The owning application’s log, which is usually more specific than anything VSS says about it.
vssadmin list writersimmediately after a failure, before anything is restarted – the state is lost the moment you fix it.
When it is none of the above
Two situations look like a VSS fault and are not. The first is contention: another product taking its own snapshot in the same window, or a virtual machine being backed up by both the host and an in-guest agent. Stagger them so one requestor is active at a time and see whether the failure becomes reproducible. The second is a source volume that cannot be snapshotted at all, which fails immediately and consistently rather than under load; remove it from the set to confirm, then decide how that data is protected instead.
Re-registering VSS components is advice that dates from much older Windows versions. Restarting the owning service, removing a stale provider, fixing the destination and clearing a dirty volume resolve the large majority of these without touching component registration.
Every code this article covers
| Code | What it points at | Source |
|---|---|---|
0x807800C5 |
Reported by Windows Server Backup when the shadow copy stage does not produce a usable snapshot. Microsoft publishes no meaning for the code, so work from the numeric code inside Event ID 521 | not published by the vendor |
Event ID 521 |
Backup started at the stated time failed because the Volume Shadow copy operation failed for the backup volumes, with the underlying error code printed in the event | Microsoft Learn |
Event ID 517 |
Referenced by Microsoft as a Windows Server Backup failure entry in Event Viewer, but with no published message text. Read it alongside 521 | not published by the vendor |
0x8078015B |
Windows Backup encountered an error accessing the remote shared folder used as the target; retry once the share is available and accessible. For latency, SessTimeout under LanmanWorkstation\Parameters is the documented adjustment | Microsoft Learn |
Event ID 12293 |
Microsoft’s backup guidance treats a 12293 in a guest, alongside 8229 on the host, as indicating VSS writer errors. No message text is published for the ID | Microsoft Learn |
Confirm the fix worked
vssadmin list writersshows every writer in [1] Stable with no last error.- A full backup job completes and records success.
- The job’s detail log carries no per-volume warnings.
vssadmin list shadowstorageshows every source volume with an association that has room in it.- A test restore of a single file from the new backup proves the image is readable, not merely written.
Questions people ask about this
Does fixing this require buying backup software?
No. Windows Server Backup is included with Windows Server, and the failure is in the snapshot stage – which a different backup product would hit in exactly the same way, because it uses the same VSS plumbing.
Is the target really worth checking?
Yes. A stale WindowsImageBackup folder, a nearly full or non-NTFS target, a shadow copy sitting on the target volume, or a failing external disk all produce this, and all four are quicker to check than any writer investigation.
Why does the job sometimes succeed on a retry?
Because the failing condition is often transient: a writer that was busy, a volume momentarily short of space, or two requestors colliding. A backup that succeeds only sometimes is still a backup you cannot rely on.
Can I just exclude the volume that fails?
For a genuinely disposable data volume, possibly. Never for a system volume: you will discover at restore time that you have a backup which cannot bring the server back.
Should I re-register the VSS components?
Not first, and usually not at all. That advice predates current Windows Server versions. Work through the destination, the writer states, the provider list and the source volumes instead.
