I been running my Unraid system with a Windows 11 VM and a Debian 12 VM.

Periodically, I find the Unraid system has crashed and restarted, and the Windows VM is stuck at "Repair Windows"

I repeatedly restart the Windows VM from the VNC console. The first time it again crashes Unraid and restarts, coming to the same boot failure for Windows.

Eventually, Windows starts to boot, and when it arrives at the desktop, there is a notification that Windows Update failed and was removed.

How can I prevent these WIndows updates and the resulting complete system failure of Unraid and Windows that they are causing ?

Edit: Suggestion for Unraid VM improvement - The VM status shows "Windows 11- Started", even if boot has failed. It would be helpful to modify the status also show something like "Boot Failed", or "Pre-boot" instead of "Started" until Windows has actually successfully booted.

Edited July 9, 2025Jul 9 by timg11

Maybe I'm the only one using Windows 11 in a VM under Unraid?

My workaround is using GPEDIT.MSC to "Configure Automatic Updates" to option 3 "Auto Download and notify for install".

That won't help with updates rendering the VM unbootable, but at least I can do updates when I'm present and ready to address the issue.

For the Unraid VM experts here, apparently Windows tries a few reboots to see if it can auto-repair. However, it appears that Windows in a VM gets confused, and restarts the entire machine, not only the VM. That causes an unclean restart of UnRaid, and days of parity reconstruction. Are there any settings in the VM that could "isolate" the Windows restart attempt from affecting the underlying hardware and the Unraid OS?

Edited July 13, 2025Jul 13 by timg11

I've had this WIndows 11 VM running in Unraid for a few months.

Twice it has failed during updates and not only failed to re-start Windows 11, but brought down the entire Unraid system.

The lack of response to this topic has me wondering which of the following is true:

1) I am one of very few people using Windows 11 in a VM under Unraid, so nobody has any knowledge of issues.

2) Others are using a Windows VM in Unraid, but I am one of very few people experiencing failures of WIndows Updates which crash the entire Unraid system.

Please reply here if you are successfully running Windows 11 as VM in Unraid, and your Windows Updates have been installed without any issues. I'd like to compare notes on the system and environments.

I Run a Windows 11 VM but its only started 3 or four times a month for a few hours at a time. I manually install windows updates every patch Tuesday and have never had any issues. Running latest version of Unraid.

I have two continuously running Windows 11 VMs on Unraid 7.0.1, both install Windows Updates automatically or manually without issues. Happy to pass along configs if it might help, but I don't recall doing anything special in the setup of either.

running win11 VM's, with or without passthrough on several hardware and Unraid Servers ;)

no issues here ...

Good to know that others are successful with a Windows 11 VM.

So what steps can I take to understand why anytime the Windows VM restarts (manually or for updates), it unexpectedly restarts the host Unraid system (followed by 24 hours of parity check). That is a restart from within the Windows VM. If I use the Unraid VM control to stop or restart, it works properly.

Are there logs that might give a clue? I look at the Unraid system log, but it starts at the time the system restarts after the unclean shutdown. Are older logs kept anywhere?

I have a USB3 interface and a VGA video passed through to the Windows VM. Would either of those cause a VM restart to bring down the host?

Edited August 30, 2025Aug 30 by timg11

13 minutes ago, timg11 said:

Are older logs kept anywhere?

The syslog in the diagnostics is the RAM copy and only shows what happened since the reboot.   It could be worth enabling the syslog server to get a log that survives a reboot so we can see what happened prior to the reboot. The mirror to flash option is the easiest to set up, but if you are worried about excessive wear on the flash drive you can put your server’s address into the Remote Server field.

34 minutes ago, itimpi said:

you can put your server’s address into the Remote Server field.

My Unraid system is my server, so I don't think that would work. I think you mean another server. I have a Raspberry Pi on the network, so I configured it as a syslog server and pointed Unraid's remote syslog to the Pi's IP. We'll see what that log file shows next time.

2 minutes ago, timg11 said:

My Unraid system is my server, so I don't think that would work

It DOES work! It is mentioned in the syslog server link.

That is Interesting - I don't deny it would work, but I wonder about the case of the system being restarted unexpectedly - a non-clean shutdown. Would the data get written to the local file?

With a remote syslog server, Unraid just has to send the UDP out. Using itself as the syslog, it has to send the UDP, receive the UDP, and write it to the log, files have to be flushed, etc. Maybe it would all get done as the system goes down, but it seems safer on the separate server that can write the log and is not in the process of going down.

6 hours ago, timg11 said:

That is Interesting - I don't deny it would work, but I wonder about the case of the system being restarted unexpectedly - a non-clean shutdown. Would the data get written to the local file?

best practice to debug in my experience

1/ a monitor will show the most results

2/ syslog to flash

3/ syslog server

now, a "hard" reboot by itself usually means, hardware failure somewhere ... if the OS crashes it usually feeeze and you have to hard reboot yourself.

so may describe, what is meant with "bring my whole unraid system down"

1/ OS freeze

2/ Server hard reboot by itself

on 1/ see upper debug choices and lets see

on 2/ ... hardware, very hard to debug ... most likely trial & error, RAM, Heat check, PSU check, CPU check, Mainboard check, ... may too aggressive active power savings ...

11 hours ago, timg11 said:

Would the data get written to the local file?

If anything I think this is more likely to work. If sending to a remote server the network has to stay operable long enough for it to get there, whereas when being written locally it should be faster. Having said that I am not sure I have seen a comparison of what is captured when you write locally as opposed to when you write remotely to give a definitive answer.

The one that is more likely to work is the "mirror to flash" option as that does not use the network.

13 hours ago, alturismo said:

so may describe, what is meant with "bring my whole unraid system down"

1/ OS freeze

2/ Server hard reboot by itself

It is option 2. The WIndows 11 VM is running or starting, and decides to do an update. Windows Updates always involve a reboot. The strange thing is that the Windows VM reboots doesn't just restart the VM, but reboots the entire T440 server running Unraid and hosting the VM.

It seems to be an expected reboot from the Windows perspective, but unexpected from Unraid.

I don't suspect hardware issues, since the server runs for weeks without issue, until the Windows VM tries to do an update. I have it configured to not update without asking, but it will still perform an update if rebooted. So the issue happens when I need to shut down the VM and it decides that is the time to do an update.

Edited August 31, 2025Aug 31 by timg11

10 hours ago, timg11 said:

I don't suspect hardware issues

i certainly do (sadly) as

1/ the VM runs on its own virtual hardware, so no access to reboot (expected or unexpected) the Host.

2/ the "hard" load from an update triggerung your hard crash.

may post your diagnostics to take a closer look, at least try to look for

memtest, power saving features (disable them), to check ...

I did a simple test of stopping and restarting the Windows 11 VM.

I captured a video of the result. Attached as PXL_Unraid_VM.TS.mp4. It is between about 16:36 and 16:46 local time.

When restarting the VM, the entire T440 restarts. There are some interesting patterns on the 6 LED bar on the motherboard, but it seems those are undocumented.

The VM startup fails, and it restarts, taking Unraid with it.

The second time around it restarts Unraid again, and I interact with the VNC and tell it to go ahead and start Windows, which it does.

I have attached diagnostics, taken after the video was recorded and the system is running again.

I also attached the log from the raspberry pi that I set up as a syslog server, with the portion before the reboots. t440log-09-01.log

t440log-09-01.log

t440-diagnostics-20250901-1719.zip

Edited September 1, 2025Sep 1 by timg11

6 hours ago, timg11 said:

I did a simple test of stopping and restarting the Windows 11 VM.

may asked, possible you are using "RAM ballooning" in your win VM ?

if so, set static and NOT dynamic, initial and may should be equal in size, like 16 / 16 GB

windows doesnt really like that, may that triggers something in your RAM which causes the Server to crash, usually it just likes to crash the VM, but with may bad addressing, bad RAM, bad ... it may be the cause of your issues.

also, about this

On 8/30/2025 at 9:06 PM, timg11 said:

I have a USB3 interface and a VGA video passed through to the Windows VM. Would either of those cause a VM restart to bring down the host?

when i see your vfio binds, they are empty ... some hardware doesnt like to be passed and "resetted" ...

and what VGA are you passing ... ? that matrox chip ? may atleast also try without passing that through, also USB pass ...

attach a screen from your VM setup page where its clear which devices you inserted and how ...

@alturismo thanks for the reply. I'm going to be unable to work on this for the next two weeks, but I will get back to it.

From what I can tell, I can't look at the VM configurations while the VM is running. I don't want to stop it until I have a plan of action and am ready and able to spend time with it.

In the meantime, please tell me more (or point me to details) on "RAM ballooning". I'm not familiar with it, and would not have enabled it unless it is a default.

Yes, the Matrox is the only video built into the T440 system. Please explain what you mean by "when i see your vfio binds, they are empty". The binding was one of the first things I had to figure out. I had a thread on the VGA passthrough and another thread on the USB controller passthrough. These show some of the settings, which have not changed since I got it working.

If I can get any other information about the VM setup without stopping the VM, let me know, otherwise I'll stop it and share the detailed setup in a couple weeks.

I have ordered a Nvidia Quadro NVS 300 video card with PCIe X1 that I will install at that time. I will remove the Matrox and pass through the Nvidia to Windows. That will allow the Matrox to remain dedicated to the Bios and the Unraid shell. The Nvidia Quadro will be used only by the Windows VM, so hopefully reduce any problematic "resetting" at VM startup or shutdown.

1 hour ago, timg11 said:

I can't look at the VM configurations while the VM is running.

of course you can, just dont edit stuff while its running ;)

1 hour ago, timg11 said:

In the meantime, please tell me more (or point me to details) on "RAM ballooning". I'm not familiar with it, and would not have enabled it unless it is a default.

those should be equal (here sample from a running VM)

ballooning would be like, start with 8 GB, go up to 24 GB as needed

1 hour ago, timg11 said:

I had a thread on the VGA passthrough and another thread on the USB controller passthrough. These show some of the settings, which have not changed since I got it working.

when i see this (also no entries in syslinux ...

but yes, vfio-pci is in use it seems, but i wouldnt try to pass the matrox onboard to a VM ... also sadly often critical, also when i see

its in a group with a pcie bridge, passthrough's should always be "complete" group or a single iommu device, sample

may another screen from the tools, devices page where those devices are visible.

OK, here are some details from the VM "Edit" settings while it is running.

So yes, the Max memory is > Initial. So when I edit the VM, I will set them both to 16 GB so there will be no changes.

It looks like there is enough free RAM to support that:

You said "passthrough's should always be "complete" group or a single iommu device, sample"

I'm not sure where the settings like your sample are found? Maybe they are not visible when the VM is running? Or I've forgotten where the IOMMU settings and XML files are reached in the GUI?

Hopefully when I remove passthrough for the Matrox VGA and fully passthrough the new Nvidia video, that will resolve the issues.

On 9/7/2025 at 7:14 PM, timg11 said:

So yes, the Max memory is > Initial. So when I edit the VM, I will set them both to 16 GB so there will be no changes.

It looks like there is enough free RAM to support that:

yes, you have "ballooning" active which is definately not a good idea on windows (hot extend RAM ...) ;)

about the matrox ... you have this as secondary ... shouldnt really be a issue, but once you switch, put the Nvidia as primary ;)

dont forget, either connect it to a monitor OR use atleast a dummy (DP, HDMI, ...) to get it proper initialised.

On 9/7/2025 at 7:14 PM, timg11 said:

You said "passthrough's should always be "complete" group or a single iommu device, sample"

I'm not sure where the settings like your sample are found? Maybe they are not visible when the VM is running? Or I've forgotten where the IOMMU settings and XML files are reached in the GUI?

thats not on the VM settings, thats on tools, system devices ...

I am back to trying to solve this problem. I have set the memory size to 16384 for Min and Max, so that should stop balooning.

However, the VM is still crashing and bringing down the host with every restart. When I start with VNC I can see the Unraid logo with the Windows spinning circle, which freezes, then the Unraid server restarts.

I then disabled the Matrox passthrough in Tools / System Devices by removing the XML that I added earlier:

    <hostdev mode='subsystem' type='pci' managed='yes'>

      <driver name='vfio'/>

      <source>

        <address domain='0x0000' bus='0x03' slot='0x0' function='0x0'/>

      </source>

    </hostdev>

Now I can shut down and restart the Windows VM and Unraid keeps running.

I should be able to shut down a Windows VM either by using the Stop button in the Unraid VM control page, or from within the VM by using Start Menu / Shut Down.

Are they completely equivalent? Either way better (safer) than the other?

Also I notice a lot of problems with the Unraid VNC, where the mouse doesn't work right, I can't access Start menu, etc. It works fine from Keyboard so OK for BIOS and startup settings. Is this a known issue?

I have a TightVNC server running in the VM, and I use the TightVNC viewer for day to day work, but the Unraid VNC is needed for startup tasks.

Edited September 27, 2025Sep 27 by timg11

Now on to making the NVIDIA NVS 300 card available to the Windows VM.

Here are the details from Tools / System Devices.

There is the option to click the checkboxes in that menu and click on the button "Bind Selected to VFIO at boot"

Or there is the XML way, which would be presumably to add this to the VM configuration:

    <hostdev mode='subsystem' type='pci' managed='yes'>

      <driver name='vfio'/>

      <source>

        <address domain='0x0000' bus='0x17' slot='0x0' function='0x0'/>

      </source>

    </hostdev>

Are these doing the same thing, or is there a benefit of one over the other?

Edit1: checking the checkboxes for "Bind Selected to VFIO at boot" does not work. The NVIDIA is not visible to the Windows VM. The video is not initialized with a monitor connected.

So I assume the XML is required.

Unfortunately, the checkbox method in System Devices requires a full reboot of Unraid, so I'll have to reboot again.

I'm discovering that the "Reboot" button on the Main page of Unraid doesn't seem to work right. It shuts down, but the system never came up again after 10 minutes. I had to use IDRAC to do a hard shutdown of the server, then Unraid started again.

When I unchecked the VFIO bindings, I just used Shutdown directly from the Main page.

Edited September 27, 2025Sep 27 by timg11

7 hours ago, timg11 said:

Now on to making the NVIDIA NVS 300 card available to the Windows VM.

there is nothing you can do with this card anyhow, may rather use none to test ?

that card is almost 20 years old ?

7 hours ago, timg11 said:

There is the option to click the checkboxes in that menu and click on the button "Bind Selected to VFIO at boot"

thats the proper way

1/ vfio bind device/s

2/ reboot

3/ attach device/s to your VM

I selected the NVIDIA NVS 300 because it fits in an PCIe 1X slot. My use case is only a basic display, no GPU, video, or anything demanding is required.

I'm having problems getting it to run - I've got the VFIO binding done, the XML for hostdev in the VM.

Device Manager recognizes there is an NVS300 in the Windows VM, but it is giving Error 43.

I will start another thread on that topic.

Thank you for your help - by removing the Matrox pass-through I am now able to start and stop the Windows VM without crashing Unraid.

The matrox VGA output now continues to display the Unraid boot terminal, and eventually goes to sleep.

I'm not sure if memory ballooning was also a problem, but eliminating that alone did not stop the crashes. I'll leave it at 16G fixed for now.