VM CPU usage increased after proxmox upgrade from pve6.4 to pve7.0

ferrets

Member
Oct 2, 2020
15
2
23
34
I've upgraded my proxmox ve from pve6.4 to pve7.0 last night and I found that the CPU usage of VMs had increased around 20%~40%, and I've total no idea what's the reason.

1.PNG2.PNG3.PNG
 
  • Like
Reactions: bofh1337
I thought I was going mad - seen exactly the same thing - my node cpu load and % cpu usage appears to have tripled on all VMs and on the node as a whole.

For reference:

CPU(s)

4 x Intel(R) Core(TM) i3-8109U CPU @ 3.00GHz (1 Socket)

Kernel Version

Linux 5.11.22-1-pve #1 SMP PVE 5.11.22-2 (Fri, 02 Jul 2021 16:22:45 +0200)

PVE Manager Version

pve-manager/7.0-8/b1dbf562

Repository Status

Proxmox VE updates Production-ready Enterprise repository enabled
 

Attachments

  • Screenshot 2021-07-11 120046.jpg
    Screenshot 2021-07-11 120046.jpg
    104.6 KB · Views: 10
  • Like
Reactions: bofh1337
im guessing the scheduler isnt tuned well to respond to moderate levels of load.

my local proxmox currently is plugged into a watt meter, so I may compare the schedulers to see actual effect, but I already have 30w saving over esxi at idle and its on the performance scheduler 55w vs 92w.
 
I have this same issue after recently upgrading from PVE6.4 to PVE7.1-7. I'm using a Xeon E-2276G.

Windows VMs are very sluggish and CPU Load is constantly high, despite scaling governor being set to "performance".
Is there a workaround for this?
 
I have the same problem since yesterday after updating to latest kernel and Qemu version. The memory integrity was already disabled and the problem is the same.
 
On a windows server 2019 vm the cpu load was on average 15% and now it is more than 60%. This machine was already on proxmox 7.1 and working fine. Yesterday i updated it with latest kernel and qemu. The scaling governor is set to perfromance also.