> Date: Tue, 13 Feb 2018 15:30:39 -0800
> From: Mike Larkin <mlar...@azathoth.net>
> 
> On Wed, Feb 14, 2018 at 11:01:00AM +1300, Megan wrote:
> > We have recently installed OpenBSD 6.2 onto the latest version of
> > Microsoft's Hyper-V.  It is all running smoothly with the exception of one
> > aspect, namely timeouts - the latter are expiring in 1/2 their prescribed
> > interval.  The following script shows the issue:
> > 
> > # while 1
> > ? sleep 1
> > ? date
> > ? end
> > Wed Feb 14 10:19:23 NZDT 2018
> > Wed Feb 14 10:19:24 NZDT 2018
> > Wed Feb 14 10:19:25 NZDT 2018
> > Wed Feb 14 10:19:25 NZDT 2018
> > Wed Feb 14 10:19:26 NZDT 2018
> > Wed Feb 14 10:19:26 NZDT 2018
> > Wed Feb 14 10:19:27 NZDT 2018
> > Wed Feb 14 10:19:28 NZDT 2018
> > 
> > The issue is also apparent with commands such as "vmstat 1" - this returns
> > two reports per second as opposed to one.
> 
> Interesting that people have reported exactly the opposite with kvm/qemu
> (timeouts taking twice as long).
> 
> I'll check a few Hyper-V servers and see what I come up with and report back.

Well, this one is probably related to:

> cpu0: Intel(R) Xeon(R) Gold 5120 CPU @ 2.20GHz, 1709.96 MHz

and/or:

> cpu0: apic clock running at 114MHz

which almost certainly means the calibration of the timecounter is wrong.

  sysctl kern.timecounter.choice=acpitimer0

might improve things.

Reply via email to