I have two windows 10 vms on a proxmox 8.4 host with exactly the same specs.
CPU: 10 cores host
Ram: 24 GB
Balloon : 8 Gb
virtio io drivers installed
connectx-4 lx virtual device passthrough
Tesla P40 virtual device passthrough
Disk options:
Virtio scsi single : no cache, ssd emulation on
OVMF bios
Randomly, the proxmox host starts reporting 90% ram usage in one of them, even though windows reports about 3 GB ram usage.
I tried updating the virtio drivers , removing the pcie passthrough, yet nothing has worked.
They both were using 248 driver, I updated to the 285 on the problematic vm but still facing the same issue
I have noticed this happening on a number of windows vms (10 , 11 , server 2022) randomly. The only solution I've found is to remove the vm entirely and create a new or an os clean install (I think)
My system spec:
Amd epyc 7531
256 gb ram
Zpool for vm storage.
Any pointers how to sort his out is going to great, since I can't do a clean install on certain important VM's
On the problematic VM:
On the good VM
UPDATE:
SOLUTION:
I could not find this anywhere on the forums, so I am writing the solution here
If there is a spec change on the mahcine, ( host processor change , disabling numa... ) can result in windows throwing PerfOS erros . This can be seen in the windows event viewer. This error is thrown when the balloon driver is unable to collect the physical memory utilization data. I have shared a screen shot of the issue
The solution worked as shown on the link below
Solution 1 (This worked for me):
A possible solution might be to rebuild the performance counter via lodctr (run inside a “run as admin” CMD):
lodctr /R
If this caused the error message:
Error: Unable to rebuild performance counter setting from system backup store, error code is 2
run it a 2nd time until it shows:
Info: Successfully rebuilt performance counter setting from system backup store
Solution 2:
Another possible solution might be to check the page file (it shouldn´t be fully disabled) on the affected server and re-adjust that if needed.
Sources and reference:
https://www.admin-enclave.com/compu...-of-the-data-section-contains-the-status-code
https://social.technet.microsoft.co...0e4cf/2012-r2-numa-warning?forum=winservergen
CPU: 10 cores host
Ram: 24 GB
Balloon : 8 Gb
virtio io drivers installed
connectx-4 lx virtual device passthrough
Tesla P40 virtual device passthrough
Disk options:
Virtio scsi single : no cache, ssd emulation on
OVMF bios
Randomly, the proxmox host starts reporting 90% ram usage in one of them, even though windows reports about 3 GB ram usage.
I tried updating the virtio drivers , removing the pcie passthrough, yet nothing has worked.
They both were using 248 driver, I updated to the 285 on the problematic vm but still facing the same issue
I have noticed this happening on a number of windows vms (10 , 11 , server 2022) randomly. The only solution I've found is to remove the vm entirely and create a new or an os clean install (I think)
My system spec:
Amd epyc 7531
256 gb ram
Zpool for vm storage.
Any pointers how to sort his out is going to great, since I can't do a clean install on certain important VM's
On the problematic VM:
Code:
qm monitor 301
qm> info balloon
balloon: actual=24576 max_mem=24576 last_update=1765571446
qm>On the good VM
Code:
qm monitor 305
qm> info balloon
balloon: actual=24576 max_mem=24576 total_mem=24554 free_mem=18794 mem_swapped_in=4888313856 mem_swapped_out=0 major_page_faults=223759 minor_page_faults=43613791 last_update=1765597543
qm>UPDATE:
SOLUTION:
I could not find this anywhere on the forums, so I am writing the solution here
If there is a spec change on the mahcine, ( host processor change , disabling numa... ) can result in windows throwing PerfOS erros . This can be seen in the windows event viewer. This error is thrown when the balloon driver is unable to collect the physical memory utilization data. I have shared a screen shot of the issue
The solution worked as shown on the link below
Solution 1 (This worked for me):
A possible solution might be to rebuild the performance counter via lodctr (run inside a “run as admin” CMD):
lodctr /R
If this caused the error message:
Error: Unable to rebuild performance counter setting from system backup store, error code is 2
run it a 2nd time until it shows:
Info: Successfully rebuilt performance counter setting from system backup store
Solution 2:
Another possible solution might be to check the page file (it shouldn´t be fully disabled) on the affected server and re-adjust that if needed.
Sources and reference:
https://www.admin-enclave.com/compu...-of-the-data-section-contains-the-status-code
https://social.technet.microsoft.co...0e4cf/2012-r2-numa-warning?forum=winservergen
Last edited:
