Re: [chrony-users] Chrony offset and stability adjustments?

[ Thread Index | Date Index | More chrony.tuxfamily.org/chrony-users Archives ]


On Mon, Jan 07, 2019 at 09:28:50AM -0800, Denny Page wrote:
> I don’t believe a guarantee of < 1 ms from true time is achievable inside a VM. Too many things happen in a VM system even when the host is not over subscribed.

I think that depends on the VM. On KVM, 1 millisecond or even better
is not a problem in my experience.

In general, if the guest doesn't migrate and can use the real TSC of
the CPU as the clocksource, its system clock will be independent from
the host's system clock and will be as stable as the HW is.

Linux network drivers (e.g. virtio and e1000) used in VMs support SW
timestamping. With the interleaved mode that works very well.

For a single-digit microsecond, or even sub-microsecond, accuracy, a
different approach is needed. The host's system clock needs to be
synchronized using HW timestamping and the guest is synchronized to
the host using a virtual PTP clock, e.g. provided by the ptp_kvm
driver.

> For a few hundred dollars you can buy a small hardware device that serves NTP (or even PTP) based on GPS. Very, very cost effective.

That's great until you hit a bug and the vendor doesn't provide the
fix. You have no source code and can't do anything about it.

I prefer open source solutions.

-- 
Miroslav Lichvar

-- 
To unsubscribe email chrony-users-request@xxxxxxxxxxxxxxxxxxxx 
with "unsubscribe" in the subject.
For help email chrony-users-request@xxxxxxxxxxxxxxxxxxxx 
with "help" in the subject.
Trouble?  Email listmaster@xxxxxxxxxxxxxxxxxxxx.


Mail converted by MHonArc 2.6.19+ http://listengine.tuxfamily.org/