Re: [chrony-users] Tracking lost but server selected

[Ariel Garcia <a.garcia@xxxxxxxxxx>]

Hello,

Sorry i had sent this email on Friday but something went wrong.
I now have a lot more of data (debug logs, chrony logs) showing the 
correlation between "reference lost" and "fallback drift activation" but i 
think the data from Friday should be quite helpful already.

----

> > you mean the measurements are too bad to enable selecting a reference
> > source faster than in several hours?
> 
> They seem too good to cause the sources to have a large skew and
> quickly reach maxdistance when they are missing a larger number of new
> measurements.
> 
> > ("maxdistance 3s" roughly means that measurements with RTT > 3s are
> > discarded right?)
> 
> maxdistance is only related to the source selection. The maxdelay*
> options are related only to the individual measurements. If we ignore
> the skew and other sources of dispersion, a source which doesn't have
> a measurement with better delay than 6 seconds would always fail the
> maxdistance check. (distance = delay / 2 + dispersion)
> 
> Do you set any of the sources to offline (with chronyc)?
> 
> Offline sources are treated as unreachable and have some additional
> requirements for selection.
> 
> It would probably help if we could see debug output from the time when
> the sources cannot be selected.

Ok, got it :)

There seems to be a 1 to 1 correlation with the "Fallback drift 13" 
activation, which is the value i selected in the config file:
	fallbackdrift 13 19

Attached the log files and all the debugging output since a device reboot
on 21st March. Another app turned the network connection down around 
6-9am on March 22nd, but then the problem appeared at 13:35 and then 
again at 18:28 on March 22nd.

Thanks again!

Attachment: logs-reference-lost.tgz
Description: application/compressed-tar