Fix it now
0x80073712 means the component store Windows services itself from is missing pieces, so updates, optional features and system file checking all fail together. Repair the store first, then let the file-level check run against a store that has something sound to copy from.
DISM /Online /Cleanup-Image /ScanHealth
DISM.exe /Online /Cleanup-image /Restorehealth
sfc /scannow
- Let RestoreHealth run to the end. It can sit at one percentage for a long time; interrupting it leaves the store in a worse state than it started.
- If it reports that the source files could not be found, supply a source. Mount Windows installation media, or use a healthy machine of the same build, and re-run it with
/Source:and/LimitAccess. - Point
/Source:at a mounted or extracted Windows folder, such as\\server\c$\windows, rather than at the .wim file itself. - Run
sfc /scannowonly after the store reports healthy, then restart and retry whatever was failing.
Run DISM before the System File Checker, not after. SFC repairs files using copies held in the component store, so a damaged store gives it nothing sound to copy from.
If ScanHealth now reports no corruption and the update installs, you are done. If the repair itself fails, the next section explains what it is failing on.
Why it happens
Windows keeps every component it is built from, in every version it has been serviced to, in a store on the system drive. Servicing does not patch files in place: it adds a new version of a component to the store and repoints the system at it. The file checker works the same way, replacing a bad file with the copy the store holds.
That makes the store a single point of failure for everything that changes the operating system. If a manifest is missing or a payload file has gone, the servicing stack cannot satisfy the request and stops rather than proceed in a half-known state. One damaged component turns into failing updates, features that will not enable, and repair commands that report they found something they cannot fix.
The three codes are not three versions of one thing, which is why the fix differs between them. 0x80073712 says the store is corrupted. 0x80073701 says a specific referenced assembly could not be found, which is a missing piece rather than a broken index. 0x80070308 is a different animal: an operation requested that is not valid for the state the component is in, which in servicing usually means an earlier transaction is still waiting on a restart. Fetching a repair source will not help that one.
An update was interrupted part way through
You have this one if the failures began after a power loss, a forced restart during servicing, or a volume that filled up mid-update.
- Make working room on the system volume, because servicing needs it:
fsutil volume diskfree c: - Run
DISM.exe /Online /Cleanup-image /Restorehealthand let it complete without interruption. - Follow it with
sfc /scannow, then restart and retry the update that was failing.
A servicing transaction is still pending
You have this one if the code you get is 0x80070308 rather than 0x80073712, or a repair reports success and then fails again identically on the next attempt.
- Restart and let the machine finish any pending servicing work first. That is what the code is asking for.
- Check the update history for something stuck part-installed, and let it finish or roll back.
- Only once it has restarted cleanly, re-run ScanHealth to see whether there is real damage underneath.
Treating this one as missing payload sends you hunting for a repair source you do not need.
The machine cannot reach a usable repair source
You have this one if RestoreHealth runs for a while and then reports that the source files could not be found.
- Mount Windows media of the same build and repeat the command with
/Source:pointing at the Windows folder inside it, plus/LimitAccess. - A UNC path to a healthy machine of the same build also works, and is often quicker to arrange.
- If the media supplies an ESD rather than a WIM, export the index first, as shown in the reference section.
The storage holding the store is failing
You have this one if corruption returns after every repair, or unrelated files on the same volume are also damaged.
- Run an online check of the volume:
chkdsk C: /scan - Check the drive’s health through its manufacturer’s tooling, and take a backup before further repair attempts.
- Replace failing storage before spending more time on servicing repairs. Repeated corruption on a healthy file system is unusual and hardware is the ordinary explanation.
Full reference
Choosing a source the servicing stack will accept
The rules here are not the ones most people assume, and getting them wrong is the usual reason a repair with a source still fails. Use RTM media rather than refreshed media: refreshed media has older file versions excluded from it, and the target may still need exactly those files. Then make sure the source is patched to at least the level of the target, because if the installed system is patched higher than the source, the repair fails for want of files the source has never seen. And include every language pack the installation supports, because a feature restored without all of its language components will not install.
In practice that means a machine several cumulative updates ahead of your media is a common failure, and the answer is to service the source rather than to hunt for different media. A working machine of the same build, reached over a UNC path to its Windows folder, sidesteps the problem entirely because it is already at the right patch level.
Getting an index out of retail media
Retail media often ships install.esd rather than install.wim. Read the indexes first so you supply the edition actually installed, then export the one you need.
Dism /Get-ImageInfo /ImageFile:D:\sources\install.esd
Dism /Export-Image /SourceImageFile:D:\sources\install.esd /SourceIndex:1 /DestinationImageFile:C:\repair\install.wim
The parameters are /Get-ImageInfo and /ImageFile:. The older /Get-WimInfo and /WimFile: spellings are not the current syntax.
One command not to run while troubleshooting
Dism.exe /online /Cleanup-Image /StartComponentCleanup /ResetBase reclaims space by removing superseded components permanently. Once it completes, the updates already installed can no longer be uninstalled, although future update packages still can be. It is a disk space measure, not a repair, and it is the wrong thing to run on a machine you may need to roll back.
Where the servicing story is written
| Location | What it holds |
|---|---|
| %SystemRoot%\Logs\CBS\CBS.log | The current servicing log, including which component the stack could not satisfy |
| %SystemRoot%\Logs\CBS\CBS.persist.log | Older entries rolled out of the current log; frequently holds the first failure |
| sfcdetails.txt | The file checker’s own findings, extracted from CBS.log with the command below |
findstr /c:"[SR]" %windir%\Logs\CBS\CBS.log >"%userprofile%\Desktop\sfcdetails.txt"
Policy that decides where repair content comes from
On managed machines the setting that governs this is Computer Configuration, Administrative Templates, System, Specify settings for optional component installation and component repair. It can name an alternate source path and it can stop the machine reaching Windows Update for repair content at all. If RestoreHealth fails on a domain-joined machine and succeeds on an identical standalone one, check this policy before checking anything else.
When the store cannot be repaired
If RestoreHealth completes but still reports the store is not repairable, and the file checker keeps finding violations it cannot fix, read CBS.log for the components that are failing and decide whether the machine is worth more time. An in-place upgrade using media of the same version rebuilds the operating system while keeping applications, settings and data, and it is the practical route back rather than a documented remedy for this specific condition. Take a full backup first and allow considerable time on slower hardware.
Do not copy a component store from another machine. The store is specific to that machine’s build, edition, language and update history, and transplanting it produces a system that cannot service itself and is much harder to recover than the one you started with.
Every code this article covers
| Code | What it points at | Source |
|---|---|---|
0x80073712 |
ERROR_SXS_COMPONENT_STORE_CORRUPT: the component store has been corrupted | Microsoft Learn |
0x80073701 |
ERROR_SXS_ASSEMBLY_MISSING: the referenced assembly could not be found | Microsoft Learn |
0x80070308 |
ERROR_REQUEST_OUT_OF_SEQUENCE: an operation was requested that is not valid for the component’s current state, typically a transaction still pending a restart | Microsoft Learn |
Confirm the fix worked
- Run DISM /Online /Cleanup-Image /ScanHealth and confirm no component store corruption is reported.
- Run sfc /scannow and confirm it completes without finding violations it cannot repair.
- Install a pending update and confirm it completes and survives the restart.
- Enable or repair an optional feature and confirm the operation no longer fails.
- Check CBS.log for fresh failures after the successful run, not only the exit message on screen.
Questions people ask about this
Do I need a new Windows licence to repair this?
No. Component store repair does not touch activation, and an in-place upgrade using media for the same version keeps the existing licence and activation state. This is a servicing repair, not a relicensing exercise.
Should I run SFC before or after DISM?
After. The file checker repairs files using copies from the component store, so if the store is damaged it has nothing sound to copy from. Repair the store first, then run the checker.
How long does RestoreHealth take?
It varies widely with disk speed and how much has to be replaced, and it can appear to stall at a percentage for a long time. Let it run. Interrupting it can leave the store worse than when you started.
My media matches my build but the repair still fails. Why?
Almost always because the machine is patched further ahead than the source. Service the source to the same cumulative level, or point the repair at a healthy machine of the same build over a UNC path, which is already at the right level.
Can I copy the store from a working machine?
No. It is specific to that machine’s build, edition, language and update history. Use that machine as a repair source over the network instead, which is the supported way to get the same benefit.
