Fix it now
The Windows Process Activation Service has given up restarting WsusPool: five worker failures inside five minutes and rapid-fail protection disables the pool. The pool is the symptom. An oversized update database and a memory ceiling sized for a much smaller WSUS are the cause.
Get-WsusServer | Invoke-WsusServerCleanup -DeclineSupersededUpdates -DeclineExpiredUpdates -CleanupObsoleteUpdates -CleanupObsoleteComputers -CleanupUnneededContentFiles -CompressUpdates
- In IIS Manager, select Application Pools and start WsusPool if it shows Stopped. The console works again for a while, which buys you time rather than fixing anything.
- Open its Advanced Settings and deal with the memory ceiling. Microsoft’s troubleshooting guidance raises Private Memory Limit (KB) from the 1843200 default to 4000000, and to 8000000 or higher on larger environments; its best-practice guidance sets it to 0, meaning no limit, on a server with the memory to back that.
- While you are in that dialog, the other published WSUS values are Queue Length 2000, Idle Time-out (minutes) 0, Ping Enabled False and Regular Time Interval (minutes) 0, replacing IIS defaults of 1000, 20, True and 1740.
- Restart the pool, then run the cleanup above. On a badly neglected server, run the switches a few at a time so one long step does not abort the rest.
- Reindex SUSDB afterwards, and let one full synchronisation and one client check-in cycle finish before deciding it is fixed.
Confirm the server has the memory to back an uncapped pool before setting the limit to 0. If it does not, the failure moves from IIS to the operating system, which is worse.
If the pool stays started through a full patch cycle, you can stop here. The next section explains why it is disabled rather than simply restarted.
Why it happens
The Windows Process Activation Service supervises every application pool. When a worker exits unexpectedly it starts a replacement and counts the failure, and rapid-fail protection decides when to stop trying. Its defaults are published and modest: enabled, a maximum of 5 failures, and a failure interval of 5 minutes. Cross those and the pool is disabled – Microsoft’s own wording for the message is that the application pool is being automatically disabled due to a series of failures in the processes serving that application pool.
From then on every request to the WSUS web services returns 503, clients stop reporting and the console will not connect. Nothing recovers on its own, which is the part that surprises people: the pool has been stopped deliberately, and it stays stopped until somebody starts it.
What kills the workers is almost always memory. WSUS serves update metadata over a web service, and the size of that metadata tracks how many updates the server still knows about. Superseded updates that were never declined stay in the catalogue and in the data built for clients, so as more clients sync at once the working set climbs, reaches the private memory limit, and the pool is recycled underneath requests in flight. Those failures land close enough together to trip rapid-fail protection, which is why these servers usually die in the morning.
The private memory limit is being enforced exactly as configured
You have this one if Worker failures arrive in bursts while clients sync, with the pool’s working set near its limit just beforehand.
- In IIS Manager open Application Pools, select WsusPool, choose Advanced Settings.
- Raise Private Memory Limit (KB) from 1843200 to 4000000, or higher on a large estate; set it to 0 only where the server has real memory headroom.
- Set Queue Length to 2000, Idle Time-out to 0, Ping Enabled to False and Regular Time Interval to 0.
- Restart the pool and watch it through a full client check-in cycle.
Setting the limit to 0 does not remove the ceiling, it makes the operating system the ceiling. That is fine on a server with headroom and dangerous on one without.
SUSDB is carrying years of superseded metadata
You have this one if The console times out on its own views, the cleanup never finishes, and synchronisation takes hours.
- Back up SUSDB first, then reindex it. Maintenance against an unindexed database is what makes the cleanup appear to hang.
- Decline superseded updates. Microsoft names keeping them beyond their useful life as the leading cause of WSUS performance problems.
- Run the cleanup, on a neglected server a few switches at a time so a single long step does not abort the rest.
- Repeat the decline and cleanup cycle until a pass completes quickly, then schedule it monthly.
The first cleanup on a server that has never had one can run for hours and may need several passes. Run it out of hours and do not cancel part way through.
The pool cannot log in to the update database
You have this one if SQL error 18456 in the Application log, and the console fails to connect even with the pool started.
- Read the state number in the SQL error log entry, which is where the diagnosis is: state 8 is a wrong password, states 11 and 12 are a valid login whose server access failed, and states 38 and 46 mean the requested database could not be found.
- Confirm which instance holds SUSDB by reading
SQLServerNameunderHKLM\Software\Microsoft\Update Services\Server\Setup. - For a Windows Internal Database, confirm the service is running and test the pipe with
sqlcmd -S np:\\.\pipe\MICROSOFT##WID\tsql\query -E -Q "SELECT name, state_desc FROM sys.databases". - Confirm SUSDB is ONLINE rather than recovering or suspect, and that the login named in the 18456 still has rights on it.
The worker is crashing rather than being recycled
You have this one if Event ID 1026 from the .NET Runtime with a stack trace, rather than the pool hitting a memory ceiling.
- Open the 1026 and read the Description and Exception Info fields; that is where the failing component is named.
- Look for an Application Error event in the same second, which names the faulting module.
- Treat it as an application fault rather than a capacity problem: more memory will not stop an unhandled exception.
- Where a third-party component is named, take the stack to its vendor.
Full reference
Reading the events before changing anything
| What the logs show | Where the fault is |
|---|---|
| Repeated worker failures, then the pool disabled | Rapid-fail protection did its job; find what killed the workers |
| A fatal communication error with the Windows Process Activation Service | Event 5011, whose published text names the pool and the process id and puts the error number in the data field |
| A worker that terminated unexpectedly with a process exit code | The published message for that case, which is not the same event as the communication error |
| SQL error 18456 in the Application log | The pool identity cannot log in to SUSDB; the state number says why |
| Event 1026 with a .NET stack trace | An unhandled exception rather than a memory or database limit |
| Clients reporting 503 against the WSUS URL | The pool is stopped right now, whatever stopped it |
Rapid-fail protection, as shipped
| Property | Default |
|---|---|
| Enabled | True |
| Maximum Failures | 5 |
| Failure Interval (minutes) | 5 |
Raising those thresholds keeps the pool running through repeated crashes rather than preventing them. It is a legitimate stopgap on a server you cannot fix today, and it should be recorded as one. The published values are worth knowing simply because they explain the behaviour: five failures in five minutes is not many when a pool is being recycled on memory during a morning check-in surge.
The WsusPool settings
| Setting | IIS default | Published WSUS value |
|---|---|---|
| Queue Length | 1000 | 2000 |
| Idle Time-out (minutes) | 20 | 0 |
| Ping Enabled | True | False |
| Private Memory Limit (KB) | 1843200 | 0 in the best-practice guidance; 4000000 or more in the troubleshooting guidance |
| Regular Time Interval (minutes) | 1740 | 0 |
Database maintenance, in the documented order
- Back up SUSDB.
- Create the custom indexes, if you are following Microsoft’s optional recommendation.
- Reindex the database.
- Decline superseded updates.
- Run the WSUS Server Cleanup.
- Schedule the whole cycle monthly, and do not synchronise during the maintenance window.
The order matters. Cleanup against an unindexed database on a large catalogue is slow enough to look hung, which is how servers end up with a half-finished cleanup and an administrator who never runs one again. In a hierarchy, decline superseded updates top-down and run the cleanup bottom-up.
Finding and reaching SUSDB
| Question | Where the answer is |
|---|---|
| Which instance holds SUSDB | SQLServerName under HKLM\Software\Microsoft\Update Services\Server\Setup |
| Is it the Windows Internal Database | That value contains ##WID or ##SSEE |
| How to connect on Windows Server 2012 and later | \\.\pipe\MICROSOFT##WID\tsql\query |
| How to connect on Windows Server 2008 R2 and earlier | \\.\pipe\MSSQL$MICROSOFT##SSEE\sql\query |
| Why a login was refused | The state number in the 18456 entry in the SQL error log |
When the settings and the cleanup are both done
- Measure the peak working set of the pool across a full patch cycle rather than on a quiet afternoon, and size the server’s memory from that.
- If WSUS shares the machine with other roles, move it. Its memory profile is spiky and it is a poor neighbour.
- Turn off classifications you do not deploy. Driver synchronisation in particular grows the catalogue faster than any cleanup schedule can shrink it.
- Check the Application log for the actual crash before touching rapid-fail protection again; the protection is a symptom every time.
When a licence is the actual fix
The settings, the cleanup and the reindex are free, and between them they fix most of these servers. The exception is a WSUS host that has been running since a much older release and cannot be given more memory. Moving the role to a new machine means that machine needs its own Windows Server licence: Standard covers two virtual machines plus one Hyper-V host per licence, Datacenter covers unlimited virtual machines plus one Hyper-V host. Worth weighing before you commit: Microsoft has published WSUS as no longer actively developed, while stating that all existing capabilities and content continue to be available for your deployments – so a rebuild is a fair moment to review what replaces it. Arco supplies Windows Server 2025 Standard and can check whether your existing agreement already covers the new host.
Every code this article covers
| Code | What it points at | Source |
|---|---|---|
Event ID 5002 |
The application pool is being automatically disabled after a series of failures in the processes serving it. Microsoft publishes that message and the rapid-fail protection defaults behind it, but does not attach the event ID to the text | not published by the vendor |
Event ID 5011 |
A process serving the application pool suffered a fatal communication error with the Windows Process Activation Service; the event names the pool and the process id and carries the error number in its data field | Microsoft Learn |
Event ID 5013 |
Logged by the Windows Process Activation Service around a worker shutting down. Microsoft attaches no message text to this ID, so read it with the pool state and the Application log | not published by the vendor |
Event ID 5009 |
The unexpected-termination case, whose published message names the pool, the process id and the process exit code. Microsoft does not attach the event ID to that text | not published by the vendor |
Event ID 18456 |
SQL Server error 18456, severity 14, LOGON_FAILED: login failed for the named user. The state number in the SQL error log is the diagnosis – 8 is a wrong password, 11 and 12 a valid login whose server access failed, 38 and 46 a database that could not be found | Microsoft Learn |
Event ID 5010 |
Logged by the Windows Process Activation Service in connection with worker health. No message text is published for this ID | not published by the vendor |
Event ID 1026 |
Source .NET Runtime: an unhandled second-chance CLR exception. Read the Description and Exception Info fields in the event, which name the failing component | Microsoft Learn |
Confirm the fix worked
- WsusPool shows Started in IIS Manager and stays started through a full synchronisation and a client check-in window.
- No new worker failure or pool-disabled entries appear in the System log across a complete patch cycle.
- The WSUS console loads its computer views without timing out.
- Two test clients run a detection cycle and both report a fresh last-contact time.
- A repeat cleanup completes quickly, which is the sign the database is back inside a manageable size.
Questions people ask about this
Do I have to buy anything to stop the pool crashing?
No. Every fix that matters here is a configuration change or a maintenance job. The only purchase this leads to is a new server, and only when the current one cannot take more memory.
Is setting Private Memory Limit to 0 safe?
On a server with enough RAM, yes, and it is the best-practice value. On a memory-starved server it turns a recoverable recycle into an unresponsive machine, which is why the troubleshooting guidance raises the limit to 4 GB instead. Add memory at the same time or set a deliberately high fixed limit.
How often should the cleanup run?
Monthly suits most sites, and it must include declining superseded updates rather than only removing obsolete computers. A server that has gone years without one may need several passes first.
The pool is disabled again an hour after I start it. What now?
Read the Application log for the crash rather than raising rapid-fail protection. Five failures in five minutes disables it by default, so the pool being disabled tells you the workers are dying – not why.
Should I still be running WSUS at all?
Microsoft has published it as no longer actively developed while confirming that existing capabilities and content remain available, so there is no need to rush. If you are rebuilding the role anyway, it is a fair moment to price the alternatives.
