On Thu, Nov 24, 2016 at 12:00:14PM +0100, Wielant, Ulrich wrote:
> > On Thu, Nov 24, 2016 at 10:25:26AM +0100, Wielant, Ulrich wrote:
> >> What do I have to do so that chrony no longer synchronizes to the non 
> >> available PPS/GPS source?
> >
> > Just wait a bit longer :). The PPS source was much better than the NTP 
> > servers, so it may take couple hours before chronyd assumes the clock has 
> > drifted so much that it should switch to the significantly worse NTP 
> > sources. It's a feature, not a bug. If chronyd switched immediately to 
> > another source, that would be bad for accuracy of the clock, e.g. when the 
> > GPS has a weak signal and PPS stops for couple of minutes at a time.
> 
> Thanks!
> Is there a parameter so that chrony switches faster although it is bad for 
> accuracy?

I think the only way to do that would be to shorten the polling
interval for the NTP sources and force them to keep only a small
number of samples with "maxsamples 4" for example. I'd not recommend
doing that though.

> > Right, there is currently no option to set stratum for refclocks. It 
> > wouldn't make much of a difference in the source selection anyway.
> 
> Is there an option so that common NTP servers are preferred and PPS/GPS is 
> taken only if there is no other server? This might sound strange (it is ;-)) 
> but this is a requirement for my current project. GPS/PPS should be used as 
> fallback only.

That is possible. You can increase the delay value for the refclocks,
maybe to 200 milliseconds, so they are assumed to be inaccurate and
won't be prefered if the NTP sources are reachable. This may bring an
opposite problem. When the NTP sources stop working, it will take a
while before chronyd switches to the PPS/GPS refclock.

-- 
Miroslav Lichvar

-- 
To unsubscribe email chrony-users-requ...@chrony.tuxfamily.org 
with "unsubscribe" in the subject.
For help email chrony-users-requ...@chrony.tuxfamily.org 
with "help" in the subject.
Trouble?  Email listmas...@chrony.tuxfamily.org.

Reply via email to