Windows 11 Pro 24H2 VM Random Freezes on XCP-NG 8.3
-
@DevilDan If the problem shows up on older Windows versions I would start suspecting a few other factors.
- Do you have any Windows PV drivers installed?
- Could you switch back to the Realtek NIC model instead of Intel e1000?
- Does it happen on all VMs at once or just one VM at a time?
- Any issues with Linux distros?
-
No issues on Linux VMs, the problem only occurs on Windows 11 24H2, 23H2 (so far) and Windows 10, all running in UEFI mode.
- No issues so far on a new Windows 10 VM running in BIOS mode.
- I'm always installing Xen Guest drivers 9.4 (we also tested without them).
- The VMs have nothing special installed, it's just a clean Windows setup.
- I don't remember if the issue appeared on both Windows versions at the same time, so I can't confirm a correlation, but it's a possibility.
- I had some "coalesce" issues on another Linux VM, but that's been resolved, and it doesn't seem related.
- I already tested both Intel and Realtek NIC models, same issue. However, I'll switch everything back to Realtek just in case.
- Currently, I have a Windows 11 24H2 VM and a Windows 10 VM (I deleted Windows 11 23H2—it was the most stable and used less CPU than 24H2, but since I'm testing, I'll stick with 24H2).
-
@dinhngtu
I also remember having installation issues and freezes with PFsense VMs or performance slowdowns with OPNSense in UEFI mode, I think. It feels like UEFI isn't that great? -
@DevilDan
We had a similar problem when we tested version 8.3. In our case, it was related to the memory limits of the VMs.In 8.2, setting different values for Dynamic and Static memory caused a bluescreen.
In 8.3, Windows Server VMs sometimes bluescreened, while both Windows Desktop and Server VMs occasionally froze as you described.Our VMs were not fresh installs, but we resolved the issue by setting the Dynamic and Static memory limits to the same value (both min and max).
Please do try to see if it solves your problem.
-
Thanks for your feedback!
I usually always set the Dynamic and Static memory limits to the same value because I don't like auto-scaling based on load. I prefer to allocate the actual possible usage. I've attached a screenshot of my RAM configuration; since I'm not very familiar with Xen Orchestra, I just want to confirm that this setup is correct. Normally, I've set both the minimum and maximum dynamic memory to 8GB, as well as the static memory.
So far, my Windows 10 VM (BIOS) hasn’t had any issues, but the Windows 11 VM (UEFI) still seems to freeze. I’ll double-check, but it looks like the problem is really related to UEFI.
Also, I can't edit the static min memory value in XOA, and I don't know why. Is this the expected behavior?
Just to be sure, I'll set the static min memory to 8GB (in CLI on XCP-NG console) and observe the behavior, even though the documentation suggests it shouldn't make any difference.
-
@DevilDan Yeah, you cant change the static min in xoa, it has to be done through the cli
xe vm-memory-limits-set static-min=8GiB static-max=8GiB dynamic-min=8GiB dynamic-max=8GiB uuid=xxxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxxx
This is what i did, but like you said, the documentation says i makes no difference.
-
Thanks! I changed this parameter yesterday, and everything is set to the exact same value down to the byte.
No bugs so far, but I did notice some CPU usage differences between Windows 10, Windows 11 23H2, and Windows 11 24H2. Only Windows 11 24H2 constantly consumes more CPU, staying between 6-10%, while the other versions peak between 0-3%. 24H2 is really bad...
That being said, the freezes don’t seem related to this. Let’s see if changing the static min RAM makes a difference. Would be great if it does! We’ll adjust accordingly.
-
@DevilDan No bugs for the moment, PS : I choose the "other-media-install" template with all VM, maybe the static min memory is impacted finally ?
EDIT : 20 MARCH 2025 20H45 : always NO PROBLEM
-
Console is... Frozen again..... I Think it's UEFI problem, Win 10 BIOS MODE no problem....
-
@DevilDan Could you confirm that it's UEFI mode that's causing the problem?
-
@dinhngtu I had freeze on Win 10 UEFI mode, not in BIOS mode, I have only this example
24H2 UEFI Bug, 23H2 UEFI Bug, Win 10 Uefi Bug, Windows 10 BIOS OK
-
@DevilDan I noticed that in your earlier freezes, the CPU went to 0% but in these recent freezes the CPU is pegged to 100% instead. Does your guest still behave the same way during the freeze (no RDP/XOA console but still responding to ping)? It's an advanced process, but you can try attaching a kernel debugger and see what's happening. https://docs.xcp-ng.org/troubleshooting/windows-pv-tools/#connecting-to-guests-using-serial-console
-
@dinhngtu You're right, in the first freezes, the CPU would drop to 0%. Now, in recent cases, one freeze spikes the CPU to 100% and then it drops, while another stays at 100%...
The VM still responds to ping for a while, but it eventually stops sometimes hours later. The console screen remains frozen on the Windows lock screen.
I'll look into your procedure and see if I can try it.
-
@DevilDan I see, we have to disable secure boot, but it is mandatory for Win 11 NO ?
-
@DevilDan It's only mandatory for the installation. Once it's installed you can disable secure boot and enable debugging.
-
-
-
@DevilDan it boot after re-enabled secure boot, i set 2 cpu not 4 like before
-
@DevilDan Windows 11 never fails to surprise... Can you try with Server 2025?
There's a specific timing needed to attach the debugger, refer to this thread: https://xcp-ng.org/forum/topic/10284/vm-failing-to-reboot
-
@dinhngtu Hi everyone,
Thank you all so much for your help and suggestions. After several tests, I’ve decided to stop troubleshooting and stick with Windows 10 in BIOS mode. UEFI doesn’t work properly for me, whether with Windows 10 or 11, but everything runs fine in BIOS mode.
I really appreciate the time and support from this great community!