I was wondering how I could go about giving a virtual machine somewhere around 70% of all available cpu power and a good amount of ram. How can I do this?
That’s currently not possible. LXD lets you either specify a simple number of vcpu which will get scheduled in whatever way the kernel feels like or you can pin to specific CPU cores.
I don’t see any major reason why we couldn’t apply limits.cpu.allowance
to VMs as well, though the exact behavior may be a bit random as the restriction would apply to all threads of the qemu process and possibly impact I/O as well as CPU.