About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.
Event 41 from Microsoft-Windows-Kernel-Power, "The system has rebooted without cleanly shutting down first", is written at the next start. It records that the last shutdown was not clean. It does not say why, so read its data fields and the events around it. A non-zero bug check code points to a Stop error with a memory dump to read. Microsoft documents a non-zero PowerButtonTimestamp for a restart caused by holding the power button. Zeros in both point to lost power, a hard hang, forced restart by a management layer, or a dump that could not be written.
Event 6008 is logged by the Event Log service and gives the time of the unexpected shutdown. Event 1074 is the record of a planned one.
What the error means
Microsoft's page says Event ID 41 "indicates that some unexpected activity prevented Windows from shutting down correctly" and that it might be caused "by an interruption in the power supply or by a Stop error". If Windows can record error codes as it shuts down, it checks for them in the kernel phase of the next start and adds them to the event data.
The related events, from Microsoft's Windows Server page on unexpected reboots:
| Event | Source | What it says |
|---|---|---|
| 13 | Kernel-General | The operating system is shutting down at system time |
| 41 | Kernel-Power | The system has rebooted without cleanly shutting down first. This error could be caused if the system stopped responding, crashed, or lost power unexpectedly. |
| 1074 | User32 | The process named in the event has initiated the restart on behalf of a user for a stated reason, with a reason code |
| 6008 | EventLog | The previous system shutdown at a stated time was unexpected |
| 1001 | WER-SystemErrorReporting | The computer has rebooted from a bugcheck, with the code and where a dump was saved |
Microsoft states that a normal reboot shows Event 1074 followed by 13 and 6009, and that an unexpected reboot is denoted by 41 and 6008. "With an unexpected reboot, there usually isn't an Event ID 1074 log entry." The same page says event numbers can belong to different sources, so match on source as well.
The event 41 data holds BugcheckCode, four parameters, SleepInProgress and PowerButtonTimestamp. The bug check code is written in decimal. Microsoft's example, 159, converts to 0x0000009f, using Calculator in Programmer view.
Likely causes
| Cause | What you see | How to confirm |
|---|---|---|
| Stop error (bug check) | Event 41 with a non-zero BugcheckCode, event 1001, event 6008 | The code and the dump path in event 1001 |
| Forced restart with the power button | PowerButtonTimestamp is non-zero, bug check code zero | Ask who was at the console or the remote management session |
| Power loss, hard hang, hardware fault | Event 41 or 6008 with all values zero, no 1001 | Hardware checks: power supply, memory, heat, overclocking |
| Restart by a management layer | Zeros, no 1074 | Hyper-V heartbeat, server-recovery software, or the VMware heartbeat feature, per Microsoft's list |
| A Stop error whose code and dump were not saved | Zeros, plus volmgr event 46 "Crash dump initialization failed!" | Page file and dump settings |
| Planned restart, not a fault | 1074 then 13 and 6009 | The process, user and reason code in event 1074 |
Prove the cause before changing anything
These checks only read.
1. Build the timeline. PowerShell 5.1 documents LogName, ID and StartTime as hashtable keys, and ID takes a list:
$Date = (Get-Date).AddDays(-7)
Get-WinEvent -FilterHashtable @{ LogName='System'; Id=12,13,41,1001,1074,6005,6006,6008,6009; StartTime=$Date } |
Sort-Object TimeCreated | Format-List TimeCreated, Id, ProviderName, Message
Read the provider name for each line. Microsoft notes the time shown is adjusted to the system's time zone, so check the server's zone. Look for the 1074 that explains a planned restart. Look for 41, 1001 and 6008 together for a Stop error. The page also says to check for any suspicious event before the shutdown time given in 6008, in both the Application and System logs.
2. Read event 41's data. In Event Viewer, open the event's Details tab. Record BugcheckCode, the four parameters and PowerButtonTimestamp. Convert the code to hexadecimal, padded to eight digits. Microsoft's Bug Check Code Reference is the next stop for a code; the event 41 page links to it.
3. Read event 1074. Its text names the process, the user, a reason and a reason code, such as Reason Code: 0x80020010 with Operating System: Service pack (Planned).
4. Check the dump settings. Microsoft's page gives the registry location and values:
reg query HKLM\System\CurrentControlSet\Control\CrashControl
CrashDumpEnabled 0x1 is a complete dump, 0x2 kernel, 0x3 small, 0x7 automatic. DumpFile defaults to %SystemRoot%\Memory.dmp and MinidumpDir to %SystemRoot%\Minidump. A complete dump needs a paging file on a local volume large enough for all physical RAM plus 257 MB, and a small dump needs a paging file of at least 2 MB on the boot volume.
5. If event 41 shows zeros, look for event 46. Microsoft says to check for a volmgr event 46 and then the page file configuration.
6. Read the dump. The small memory dump page gives the debugger form windbg -y SymbolPath -i ImagePath -z DumpFilePath, with srv*C:\Symbols*https://msdl.microsoft.com/download/symbols as a symbol path. Inside the debugger, !analyze -show shows the Stop error code and parameters, !analyze -v shows verbose output, and lm N T lists loaded modules. The page also names the Dump Check Utility, Dumpchk.exe, for confirming that a dump was created correctly.
7. Look at recent changes. Microsoft's page says a bug check may follow a recent driver install, an application install or an update. Compare event 7045 (a service installed) and event 19 (an update installed) with the first unexpected restart.
Fix
Stop error. Act on the code and the dump. If the timeline shows a driver installed shortly before, Microsoft says you can remove or roll back that driver and see whether the Stop error stops. Disruptive step: removing or rolling back a driver changes how hardware works and needs a restart. To back out, reinstall the driver. If the dump does not name a cause, collect a full dump and analyse it with the debugger.
Bug check shown on screen but not recorded in event 41. Microsoft says to clear the Automatically restart check box under Startup and Recovery settings. Disruptive step: the server then stays on the Stop screen until someone restarts it, so an unattended or hosted server stays down. To back out, tick the box again.
Dump not written. For a small memory dump, Microsoft says the boot volume needs a paging file of at least 2 MB; for a complete dump the paging file must be on a local volume and large enough for the dump. Fix the page file for the chosen dump type, then confirm the dump type. Disruptive step: page file and dump type changes need a restart ("You must restart Windows in order for your changes to take effect"). To back out, restore the earlier size and type. Page file sizing is its own subject; this fix only needs it to meet the dump type's minimum.
Hardware. Microsoft lists these checks: disable overclocking, run a memory checker, confirm the power supply has enough wattage for the installed devices, check for overheating, and consider a UPS if power was interrupted. For a physical server, server-recovery software may have restarted it. These are hardware actions; on a hosted server, ask the provider for host-side events.
Management layer. Microsoft says a Hyper-V VM outside a cluster can be restarted by the Hyper-V heartbeat feature when the host stops receiving a heartbeat from it, and that a clustered Hyper-V setup relates to the heartbeat-monitoring option for the VM. A VMware VM could be restarted by that platform's heartbeat feature. Find out which applies before changing it. Risk: disabling a heartbeat restart leaves a hung VM down until someone acts. To back out, re-enable the option.
Forced restart. If PowerButtonTimestamp is non-zero, the fix is process: why was the server unresponsive?
Confirm the fix
After the change, watch the timeline for the same symptoms. If the cause was a Stop error, confirm the next restart, planned or not, shows no event 41. If the dump settings were wrong, Microsoft's memory dump options page refers to a separate Windows feature for generating a dump by keyboard for testing. Risk: that is a deliberate crash. It restarts the server and has no back-out, so only do it in a maintenance window on a server that can be restarted.
Example (illustrative): file server fs01 in the CONTOSO domain restarts at 03:05 on 14 November. At the next start it logs event 41 with BugcheckCode 159, event 6008 stating the previous shutdown at 03:05:41 was unexpected, and event 1001 saying a dump was saved in C:\Windows\MEMORY.DMP. No event 1074 precedes it. The decimal 159 is 9 × 16 + 15, which is 0x9F. Event 7045 two days earlier shows a storage filter driver service was installed. The administrator copies the dump off the server, reads it with !analyze -v, and rolls the driver back in the next window.
What to watch afterwards
- Alert on 41, 6008 and 1001 appearing together, and on 41 with zeros.
- Copy
MEMORY.DMPoff the server after each Stop error. Microsoft says a second complete or kernel dump replaces the previous one when the overwrite setting applies. - Keep the reason code on every planned restart so event 1074 stays useful.
- Record power-supply and temperature checks for physical hosts.
Read next
Related articles you may want next.
- HostingPage file sizing on Windows Server application hosts: commit limit and crash dumpsHow big the page file on a Windows server running business applications should be, why there is no single rule such as 1.5 times RAM, why removing it can cause freezes and lost crash dumps, and which two counters tell you the right size.Read article
- HostingWindows service recovery settings: automatic restart without hiding the faultWhat the Recovery tab of a Windows service really does: how the failure count and reset period work, why a clean exit with an error may not trigger a restart, why a queued restart cannot be cancelled, and how to set restarts that keep a business application up without hiding a recurring fault.Read article
- HostingWhen the provider's dashboard and your server disagree: hypervisor and in-guest performance figuresWhy Task Manager inside a hosted Windows server and the hosting provider's dashboard can show different CPU and memory figures, which hypervisor measurements reveal waiting that the guest cannot see, and how to bring both views together when diagnosing a slow server.Read article
Risks and limits
These commands have not been run or lab-tested.
Microsoft publishes the event 41 page under Windows Client, and describes small memory dumps in the Windows Client and Windows Server pages with different sizes, so no size is stated here. The pages do not show the output of the commands above, so check them against your own log. Microsoft does not give a way to tell a power loss from a hard hang from the event log alone. Microsoft's event 41 page describes clearing the DirtyShutdown, LastAliveStamp and TimeStampInterval values under the Reliability registry key to stop the shutdown prompt. As advice rather than a Microsoft step, leave them alone until you have the cause, because they are the system's own record. This topic does not interpret individual bug check codes.
Sources
- Event ID 41 The system has rebooted without cleanly shutting down first — Microsoft Learn, page dated 2026-02-12, accessed 18 November 2026
- Troubleshoot unexpected reboots using system event logs — Microsoft Learn, page dated 2026-02-12, accessed 18 November 2026
- Read small memory dump files — Microsoft Learn, page dated 2026-02-12, accessed 18 November 2026
- Memory dump file options — Microsoft Learn, page dated 2026-02-12, accessed 18 November 2026
- Get-WinEvent (Microsoft.PowerShell.Diagnostics), PowerShell 5.1 — Microsoft Learn, date not shown, accessed 18 November 2026
This article was originally published on .