Fix it now
The Service Control Manager events say a service exited; they do not say why. None of the five codes in this family is published by Microsoft, so the diagnosis comes from what your own entries name, from Test-ServiceHealth, and from the setup log if the failures started during an update.
Test-ServiceHealth
Get-Service *Exchange* | Where-Object {$_.Status -ne 'Running' -and $_.StartType -like 'Automatic*'}
Get-TransportAgent
- Filter the Application log to source MSExchange Common and read the crash report entry in full. Note the process, the build and the component it names; those fields, not the event number, are the diagnosis.
- Check free space on the system, database, log and queue volumes before anything else. A full volume produces failures that read exactly like code faults.
- Start the foundation first. Microsoft Exchange Active Directory Topology underpins nearly everything, and its failure produces a cascade of service-exit entries for downstream victims.
- If the crashes began during or just after an update, treat the build as half applied and read C:\ExchangeSetupLogs\ExchangeSetup.log from the end backwards.
- Once the cause is dealt with, start the remaining services and re-run Test-ServiceHealth.
Do not decode the exit codes and stop there. Microsoft publishes no meaning for any of the five codes this article covers, so a number on its own is not a finding.
If Test-ServiceHealth reports nothing missing and mail is moving, stop here. If not, the next section covers the four things that actually stop these services.
Why it happens
There are three different reporters involved and it is worth keeping them apart. The Service Control Manager entries are Windows describing a death it did not witness: one records that a service exited and returned a service-specific code, the other that a service terminated unexpectedly and how many times. Exchange writes its own report separately, naming the process, the build and the failing component. That report is the useful one, and it is the reason to filter on MSExchange Common before anything else.
What none of them gives you is a published meaning. Microsoft does not document the Service Control Manager entries in this pair, does not document its own crash report event, does not document the performance counter event that often appears beside them, and does not publish the structured exception code that shows up in the report. Microsoft does document that the .NET runtime’s HRESULT constants use the 0x8013 prefix, which is a different thing and does not extend to this exception code. So the number tells you where to look, and the entry text tells you what happened.
There is one product fact that changes the shape of this problem in 2026. Exchange Server 2016 and 2019 went out of support on 14 October 2025, and Exchange Server Subscription Edition released on 1 July 2025 is serviced with Security Updates and Hotfix Updates rather than cumulative updates. If the crashes started after patching, which kind of update you applied determines what re-running setup will even do.
Active Directory Topology cannot find a domain controller
You have this one if The topology service is among the failures, and directory access complaints appear in the Application log from the same moment.
- Confirm the server can locate a writable domain controller:
nltest /dsgetdc:<yourdomain>. - Check the server’s DNS settings point at domain controllers running DNS, with no public resolver added as a fallback.
- Compare clocks with
w32tm /query /status. Kerberos failures against a drifted clock look exactly like a directory outage. - Start the topology service on its own, let it settle, then start the rest.
The build is half applied after an interrupted update
You have this one if The failures date from an update, and the build named in the crash report is not the one you believe you are running.
- Read the installed build:
Get-Command ExSetup.exe | ForEach-Object {$_.FileVersionInfo}and compare it with Microsoft’s build number reference. - Open
C:\ExchangeSetupLogs\ExchangeSetup.logand search backwards for the last error. The task named immediately before it is the one that failed. - Re-run the update from an elevated Command Prompt so it can resume. For a cumulative update that is
Setup.exe /IAcceptExchangeServerLicenseTerms_DiagnosticDataON /Mode:Upgrade; a Security Update or Hotfix Update is its own package and must also be launched elevated. - Let it finish, then reboot. Microsoft requires a restart after an update so that registry and operating system changes can be made.
A transport agent stops the transport service
You have this one if Only the Transport or Front End Transport service is down, and the crash report names a transport or filtering component.
- List what is loaded into the pipeline:
Get-TransportAgent. - Disable any third-party agent with
Disable-TransportAgent -Identity <name>, restart the transport service and see whether it stays up. - Third-party agents must be removed before an Exchange update and reinstalled for the new build. Microsoft names incompatible transport agents, particularly after an update, as a documented cause of transport refusing service.
A volume is full, or a security product is in the files
You have this one if Services die under load or during the backup window, and resource warnings precede them.
- Check free space on every Exchange volume, including the transport queue database volume and the logging directories.
- Apply Microsoft’s published exclusions: %ExchangeInstallPath%Mailbox, TransportRoles\Data\Queue, Logging, TransportRoles\Logs, TransportRoles\Data\SenderReputation and TransportRoles\Data\Temp, plus the Exchange processes.
- Run
fltmc filtersto find leftover filter drivers from products removed badly.
Microsoft states the risk plainly: a program such as antivirus locking or quarantining an open log or database file that Exchange needs to modify can cause severe problems including potential data loss.
Full reference
What each entry contributes, and what it does not
| Entry | Reporter | Published by Microsoft |
|---|---|---|
| Service exited with a service-specific code | Service Control Manager | No |
| Service terminated unexpectedly, with a count | Service Control Manager | No |
| Exchange crash report naming process, build and component | MSExchange Common | No |
| Performance counter entry | MSExchange Common | No |
| Structured exception code in the report | The Windows exception record | No |
This is not a gap you can close by searching harder. It is the reason the working method here is to read the entry, correlate the failures by time, and treat the codes as labels rather than as answers.
Never delete Exchange transaction log files by hand to free space. They are the only route back to a consistent database after a crash, and deleting them converts a short recovery into a restore from backup. Fix log truncation instead.
Narrowing it down before restarting anything
| What you observe | Where to look |
|---|---|
| Many services down at once, topology among them | Directory access: the topology service cannot reach a suitable domain controller |
| One service down and the crash report names one component | A managed fault inside that component |
| Failures began during an update | A half-applied build. Read the setup log and finish the update |
| Only transport will not start | A transport agent failing to load |
| Failures cluster around the backup window | A backup or antivirus agent holding files |
Supportability, and why it matters here
Exchange Server 2016 and 2019 are out of support. Customers enrolled in the Extended Security Update programme continue to receive security updates for them; everyone else is expected to move to Exchange Server Subscription Edition. That changes what you can do about a service that crashes on a known defect: on a supported build there is a fix to install, and on an unsupported one there may not be. Establish which you are on before spending a night on a workaround.
When the server will not come back
- There is no repair install in the sense most administrators mean. The supported route for a server that cannot be recovered is to rebuild the machine and run setup in recovery mode, which reads its configuration back out of Active Directory.
- Do not uninstall Exchange to try to clean up. Microsoft states that uninstalling a newer build removes Exchange from the server completely rather than reverting it.
- Setting services to restart automatically on failure buys time for a transient fault and hides a repeating crash. Use it as a stopgap while you find the cause, not instead of finding it.
- If several services are failing, fix the bottom of the stack and re-check. Most of the entries in a cascade describe victims, not causes.
Every code this article covers
| Code | What it points at | Source |
|---|---|---|
Event ID 7024 |
No description is published by Microsoft. In practice the Service Control Manager reporting that a service exited and returned a service-specific error code; read the code and the service name from your own entry | not published by the vendor |
Event ID 7034 |
No description is published by Microsoft. The Service Control Manager reporting that a service terminated unexpectedly, with a count of how many times | not published by the vendor |
Event ID 4999 |
No description is published by Microsoft. Exchange’s own report about a failing process; the process, build and component fields in your entry are what to work from | not published by the vendor |
Event ID 106 |
No description is published by Microsoft. A performance counter entry from Exchange. Read what your entry names rather than treating the number as a diagnosis | not published by the vendor |
0xE0434352 |
Not published by Microsoft. Microsoft does document that the .NET runtime’s COR_E_ HRESULT constants use the 0x8013 prefix, which is a different range and says nothing about this exception code | not published by the vendor |
Confirm the fix worked
Test-ServiceHealthreports nothing missing for the installed roles.- No new service-exit or crash report entries have appeared since the change.
- Mail is moving:
Get-Queueshows queues draining and a test message is traced end to end. - Reboot the server once and confirm every service returns without manual intervention.
Questions people ask about this
Do I need to buy anything to fix this?
No. Service crashes are a support problem and everything you need is on the server already. The one place licensing intrudes is a build that no longer receives updates, because the fix for a known defect then does not exist for you.
What does the exception code in the crash report mean?
Microsoft does not publish a meaning for it. What is documented is that the .NET runtime’s HRESULT constants use a 0x8013 prefix, which is a separate range. Work from the component and build the report names, not from the exception code.
Should I set the services to restart automatically on failure?
You can, on the Recovery tab of the service properties, and for a transient fault it buys time. It also hides a repeating crash until something worse happens.
Is it safe to reinstall Exchange over the top?
No. Microsoft states that uninstalling a newer build removes Exchange from the server rather than reverting it. The supported route for an unrecoverable server is to rebuild the machine and run setup in recovery mode.
Why do so many services fail at once?
Because most depend on Active Directory topology and several depend on each other. One failure at the bottom of that stack produces a dozen entries all describing the same event.
