When I leave my Tor sys-VM alone for a period, it starts hogging a lot of CPU power when it’s supposed to be idle (i.e. no net traffic is supposed to be going through it). This has never happened using earlier versions.
Is anyone else having this problem?
There was a similar issue with Whonix-15-WS VMs last year where Tor Browser would start to consume a ton of CPU power after a period and wouldn’t stop until shut down. I think it would be a big coincidence if the two aren’t related in some way.
https://forum.qubes-os.org/tar-browser-high-cpu-usage-while-idle-in-whonix-ws/134
Update: Recently a GW VM started to use maximum processing power with dom0 alongside it. The GW VM was unresponsive and had to be killed, and dom0’s CPU load returned to normal after the process was killed.
I think this is not whonix problem, i’ve similiar case with sys-usb too. I need to restart vm to make it normal again.
For me this is exclusively a Whonix-16-GW problem. No other VM has behaved this way, and no other stops responding while just being a relay while dom0 starts overloading.
Since Whonix-16-GW handles private communications (things you want to handle through Tor), is in a vulnerable network position, and is community template (i.e. something that’s not officially Qubes and therefore has fewer assurances), the threshold for alarm is much lower than sys-usb and many other qubes.
It’s possible that this issue with GW is due to the VM using kernel-latest (5.13) which has led to some other VMs turning unresponsive.
dom0 overloading (stuck at max CPU usage for an extended period until VM killed) is still something entirely new though.