available hypervisors in FreeBSD

Paul Vixie paul at redbarn.org
Wed Apr 1 16:41:39 UTC 2015



Udo Rader wrote:
> ...
>
> I understand, that bhyve is native to BSD and will probably be the most
> effective. But given its relatively 'young age', is it production ready
> for (non nested) x86/amd64 linux guests?

there's no libvirt for bhyve yet, which turns some people off. (not me,
i don't use libvirt in any case.)

there's significant clock drift, even with
kern.timecounter.hardware="TSC-low" in the guests:

> ...
> Jan 26 05:38:08 guests ntpd[619]: time reset +0.223304 s
> Jan 26 06:06:22 guests ntpd[619]: time reset +0.196973 s
> Jan 26 06:34:24 guests ntpd[619]: time reset +0.200070 s
> Jan 26 07:08:28 guests ntpd[619]: time reset +0.210997 s
> Jan 26 07:36:09 guests ntpd[619]: time reset +0.205481 s
> Jan 26 08:10:04 guests ntpd[619]: time reset +0.205461 s
> Jan 26 08:39:43 guests ntpd[619]: time reset +0.175491 s
> Jan 26 09:10:29 guests ntpd[619]: time reset +0.189261 s
> Jan 26 09:44:03 guests ntpd[619]: time reset +0.164616 s
> Jan 26 10:20:25 guests ntpd[619]: time reset +0.176280 s
> Jan 26 10:56:18 guests ntpd[619]: time reset +0.161555 s
> Jan 26 11:39:53 guests ntpd[619]: time reset +0.166066 s
> Jan 26 12:31:11 guests ntpd[619]: time reset +0.142994 s
> ...

(that's much worse with the default kern.timecounter.hardware value, but
still rather absurd.)

i use bhyve in production and seems altogether ready.


-- 
Paul Vixie


More information about the freebsd-virtualization mailing list