Skip to content

Est. 2011ยทMicrosoft Partner ID 7033487ยทDelivery under 3 minยทSupport 7 days a week

Your vault is empty.

Free Fix 3010

MSI 3010 and 0x80070BC9: Setup Needs a Reboot Before It Can Complete

9 min read Updated October 5, 2026 Installers, Runtimes & App Deployment

Fix it now

3010 and 1641 are not failures. Microsoft publishes both as success codes: 3010 means a restart is required to complete the install, 1641 that the installer has already initiated one. 0x80070BC9 is the genuinely unhappy one – Win32 3017, the operation failed and a restart is required before the changes can be rolled back.

The first line installs without restarting; the second restarts when you are ready

msiexec /i "C:\path\to\package.msi" /qn /norestart
shutdown /r /t 0
  1. Restart the machine. In most cases the installation is already complete and the restart is all that is outstanding.
  2. In your deployment tool, map 0, 3010 and 1641 to success. Treating every non-zero code as a failure is what turns these into incidents.
  3. If a new install fails because an earlier one is unfinished, restart first and run nothing else until the machine is back.
  4. Where you must control the timing, suppress the restart during the install with /norestart or REBOOT=ReallySuppress and take it in a maintenance window.

3010’s published text carries a qualifier worth knowing: it does not cover installs where the ForceReboot action runs. A package that uses ForceReboot behaves differently, and a script keyed on 3010 will not see it.

If the machine restarts and the product works, you are done. If the pending state keeps coming back, the next section explains what is holding it there.

Why it happens

Windows cannot replace a file that is currently in use. When an installer needs to update such a file it queues the replacement instead, and the queue is processed early in the next boot before anything opens the file. Until that happens the machine sits in an intermediate state: the new files are staged, the old ones are still live, and a further servicing operation cannot safely begin.

That is why a second install started before the restart often fails outright rather than queueing behind the first. The servicing stack refuses to start a new transaction while an earlier one is incomplete, and the refusal is reported with a code that looks like a hard failure. 0x80070BC9 is the clearest example: it is Win32 3017, ERROR_FAIL_REBOOT_REQUIRED, published as the operation having failed with a restart required before the changes can be rolled back. Its neighbour 3018 says the same thing with the restart already started.

1604 belongs to the same family from the other direction. Microsoft publishes it as ERROR_INSTALL_SUSPEND, “Installation suspended, incomplete” – a transaction that has neither committed nor rolled back. A restart is the reliable way out, after which the package installs cleanly.

The practical problem is rarely the machine. It is that scripts and deployment tools treat any non-zero exit code as a failure, so a successful install that needs a restart is recorded as an error, retried, and retried again. Each retry can queue more file replacements, which makes the pending state more entangled than it was to start with.

The install succeeded and only needs a restart

You have this one if The application appears in the installed programs list and the only symptom is the return code.

  1. Restart at the next convenient point.
  2. Confirm afterwards that the application launches and reports the expected version.
  3. Record 3010 in your deployment tool as a success requiring a restart, not as a failure.

A script or deployment tool is misreading the code

You have this one if The software is installed and working, and the deployment is reported as failed and retried repeatedly.

  1. Add 0, 3010 and 1641 to the success codes for that application.
  2. In a shell script, test the exit code explicitly rather than assuming anything non-zero is fatal.
  3. Where the tool supports it, mark 3010 as a soft reboot so the restart happens inside a maintenance window.

Repeated retries are worse than the original problem. Each attempt can queue more file replacements and make the machine harder to bring back to a clean state.

An earlier transaction is unfinished

You have this one if 0x80070BC9 or 1604 appears, or a new install refuses to start on a machine that has just been updated.

  1. Restart and let the boot-time servicing complete before signing in to anything else.
  2. Check Windows Update and let outstanding work finish, restarting again if asked.
  3. Retry the install as the first action once the machine has settled.

Something keeps putting the machine back into a pending state

You have this one if The pending indicators return within minutes of every restart and installs keep failing the same way.

  1. Look for a management agent or update cycle that installs something on every boot, and let it finish before installing anything by hand.
  2. If updates are failing rather than completing, repair the servicing stack: DISM /Online /Cleanup-Image /RestoreHealth then sfc /scannow.
  3. Read the update history for a component that installs, fails and retries in a loop; that will hold the machine pending indefinitely.

Full reference

The codes, and how a script should treat them

Code Published meaning How to treat it
0 Success Nothing further required
3010 ERROR_SUCCESS_REBOOT_REQUIRED: a restart is required to complete the install. This message indicates success. It does not include installs where the ForceReboot action is run Success, restart pending
1641 ERROR_SUCCESS_REBOOT_INITIATED: the installer has initiated a restart. This message indicates success Success, restart already under way
1618 ERROR_INSTALL_ALREADY_RUNNING: another installation is already in progress Wait and retry rather than failing
1604 ERROR_INSTALL_SUSPEND: installation suspended, incomplete Restart, then install again
1603 ERROR_INSTALL_FAILURE: fatal error during installation A genuine failure. Read the verbose log

The two failure codes that mention a restart

HRESULT Win32 Published meaning
0x80070BC9 3017 ERROR_FAIL_REBOOT_REQUIRED The requested operation failed. A system reboot is required to roll back changes made
0x80070BCA 3018 ERROR_FAIL_REBOOT_INITIATED The requested operation failed. A system reboot has been initiated to roll back changes made

Both of these say the operation failed, which is what distinguishes them from 3010 and 1641. The restart they mention is needed to undo what was half done, not to finish it. Do not treat them as success codes in a deployment tool, however similar the wording looks.

How Windows records that a restart is due

Indicator What it means
A PendingFileRenameOperations value under the Session Manager key File replacements are queued for the next boot
A reboot-pending key under Component Based Servicing The servicing stack has work to finish at restart
A reboot-required key under the Windows Update branch An update has staged changes awaiting a restart
A restart prompt in Settings Windows Update is waiting for you
A management agent reporting a soft reboot The agent saw 3010 and is waiting for a maintenance window

Do not delete the pending file rename entries to force an installation through. Those entries complete replacements that are already half applied, and removing them can leave a machine running a mixture of old and new binaries – much harder to diagnose than a restart.

Suppressing the restart properly

For unattended deployment you generally want to control when the machine restarts rather than letting each package decide. Two mechanisms do this and they are not interchangeable:

msiexec /i package.msi /qn /norestart

msiexec /i package.msi /qn REBOOT=ReallySuppress

/norestart is a command-line switch that stops msiexec restarting the machine. REBOOT=ReallySuppress is a property that also suppresses restarts a package would otherwise trigger from inside the installation, including a ForceReboot action. Neither removes the need to restart eventually: the staged replacements only apply at boot.

When a long-standing 3010 becomes a real problem

  • Later installs and updates start failing, because the servicing stack will not begin a new transaction while one is outstanding.
  • Security updates that are reported as installed are not actually in force until the restart happens.
  • Each subsequent attempt queues more work, so the eventual restart takes longer and is more likely to surprise somebody.
  • Reporting drifts: the estate looks patched in the console and is not patched on the disk.
  • Treat a device that has carried a pending restart for weeks as a maintenance backlog item, not as a harmless message.

Every code this article covers

Code What it points at Source
3010 ERROR_SUCCESS_REBOOT_REQUIRED: the install succeeded and a restart is required to complete it, except where the ForceReboot action was run Microsoft Learn
1641 ERROR_SUCCESS_REBOOT_INITIATED: the install succeeded and the installer has already initiated a restart Microsoft Learn
0x80070BC9 Win32 3017, ERROR_FAIL_REBOOT_REQUIRED: the operation failed and a restart is required to roll back the changes made Microsoft Learn
1604 ERROR_INSTALL_SUSPEND: the installation was suspended and left incomplete Microsoft Learn

Confirm the fix worked

  1. Restart the machine and confirm the application launches and reports the expected version.
  2. Rerun the deployment and confirm the tool now records success rather than a failure.
  3. Confirm the pending-restart indicators are gone after the restart.
  4. Install a second package to confirm the machine accepts new transactions normally.
  5. Check that the deployment’s success codes now include 3010 and 1641 so this does not recur across the estate.

Questions people ask about this

Does any of this involve a purchase?

No. Restart handling is deployment behaviour, not licensing. The fix is a restart or a change to how your script interprets return codes, and neither costs anything.

Can I suppress the restart indefinitely?

You can suppress it during the install with /norestart or REBOOT=ReallySuppress, which is right for unattended deployment. What you cannot do is skip it forever: the staged file replacements only apply at boot, so the machine has to restart eventually.

What is the difference between 3010 and 1641?

Both indicate success. 3010 means the restart is your responsibility; 1641 means the installer has already started one, which matters if you were planning to run further steps immediately afterwards.

Is 0x80070BC9 also a success code?

No. It is Win32 3017 and its published text says the operation failed and a restart is required to roll back the changes made. The restart finishes an undo rather than an install, so do not map it to success in a deployment tool.

Is 3010 ever a real problem?

Only in the sense that it tells you the machine is not finished. A device that sits for weeks with a pending restart starts failing later installs and updates, so treat a long-standing 3010 as a maintenance backlog rather than as a harmless message.

Related error codes

Was this article helpful?

Your feedback helps us improve our documentation.

Related articles

Free Fix MSI 1642: The Patch Package Cannot Be Applied to This Installed Product Free Fix MSI 1402 and 1406: Could Not Open Key or Write Value During Installation Free Fix MSI 1327: Invalid Drive – Fixing Broken Shell Folder and Mapped Drive Paths Free Fix MSI 1919 and 27502: ODBC, COM+ and SQL Setup Actions Fail Late in Install
โ† Back to Knowledge Base