Set priority (CPU weight) for non-running qubes?

Hi,

I know it is possible to manipulate CPU share for a VM using

xl sched-credit2 -d <qube> -w <weight>

However, that needs the qube to be running.

What is the way (if any) to set priority/weight of a qube when it is not running or upon start?

I don’t know for sure, because I never tried, but I would explore a libxl hook, as the one described in a recent “mutually exclusive/anti-affinity” thread.

One could use GitHub - 3hhh/qubes-callbackd: Daemon to react to Qubes OS events. for that.

https://wiki.xenproject.org/wiki/Tuning_Xen_for_Performance
Maybe that page can help you?

Well, if it comes to scripting it, I can do it like:

  1. Start the qube
  2. Set weight
  3. Run software in qube

However, I was hoping for a more elegant solution, like e.g. having a preference (qvm-prefs) for that. IIUC, there is no such thing though.

1 Like

You can just add more VCPUs in the qube settings.

1 Like

You can just add more VCPUs in the qube settings.

How does that work in regards to prioritization?

What if qube A and qube B both run only one single-threaded app (as the main CPU load). How would adding/removing VCPUs change priority of A and B?