After upgrading from 25.10.3.1 to 25.10.4, I am experiencing higher than before CPU idle usage; it went up from 5% to >20%. It seems that the VMs are now using more CPU but I do not understand why. For example, in Truenas one of the VMs shows ~100% CPU usage in htop, while logged into the VM it shows 96% idle. So there seems to be larger overhead now. Can this overhead be reduced?
Not as drastic, but definitely saw this too.
I had an automatic scrub running earlier this morning. Now that this is over, the over all system CPU increase is from 5% to 15%. However, this of course depends on the processor and number of cores. Interesting is that htop shows the libvirt-qe process for a Windows VM at 130% even when the process is marked S for Sleeping.
I’ll add fuel this, 24.10 to 25.10.4, idle cpu went from 1-2% with 7 debian VM’s running, now it’s 6-10% across 16 threads.
After upgrading to 25.10.3 had a massive increase in CPU usage and system load according to the GUI charts and went on for several weeks like that. A simple server reboot reduced everything to normal, 1-2% CPU usage at idle. But then I have only one Ubuntu VM and 7 Truenas apps.
I ready tried the reboot and it did not solve the issue. I note that I did not jump version. I recently upgraded from core to scale with a fresh install of 25.10.3.1. this yielded less than 5% CPU time for 5 Windows and Linux VMs. the CPU increase happened on subsequent upgrade to 25.10.4.
I don’t know of any issues, but the kernel has been updated so its certainly possible.
It would be easier to review i someone could rollback between 25.10.3 and 25.10.4 while capturing process data (e.g. top). Is there anything that is different?
If you see something, submit a bug and provide the NAS-ticket here.