On Sat, Jul 11, 2020 at 11:44:34AM -0700, Claudio Chinicz wrote: > Hi, > > I noted that my sum total of vCPUs is larger than the actual number of vCPUs. > > Maybe Xen works at the minimum in order to keep sum total of vCPUs within > limits? > > One thing I tried was to disable VT-x and then HVMs (min and max vCPUs is the > same) would not start. > > Any ideas of how Xen manages vCPUs within the limits of the processor? > > Regards >
Interesting question, Claudio. As I understand it Xen allows you to create arbitrary number of vCPUs, unrelated to the actual number of pCPUs you have. As you allocate more vCPUs the system will start scheduling calls to the pCPUs and this may impact performance.I don't think that disabling VT-x will impact this, and as you have discovered it means that HVMs will not function. To work around the limit you may find it useful to reduce the number of vCPUs in most "ordinary" qubes to 1. You can also limit the number that dom0 uses, with the `dom0-max-vCPUs` parameter at boot. If you have processor intensive qubes, you can then try allocating specific vCPUs to that qube - this is called "pinning". In Xen this can be done with the `xl vcpu-pin` command . You can make a pin hard ,"MUST use this CPU", or soft, "PREFERS to use this CPU" By default Qubes makes all qubes hard/soft pin to all CPUs. If you think that you are being impacted by the scheduler, you can try pinning on processor intensive qubes and see if it helps. There used to be a health warning saying that pinning caused as many problems as it (may) solve, so try at your own risk. unman -- You received this message because you are subscribed to the Google Groups "qubes-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/qubes-users/20200712121407.GB922%40thirdeyesecurity.org.
