[ntp:questions] Ginormous offset and slow convergance

unruh unruh at invalid.ca
Wed Nov 30 17:15:52 UTC 2011

On 2011-11-30, Pete Ashdown <pashdown at xmission.com> wrote:
> David Woolley <david at ex.djwhome.demon.invalid> writes:
>>Richard B. Gilbert wrote:
>>> On 11/29/2011 1:42 PM, Pete Ashdown wrote:
>>>> +time-C.timefreq .ACTS.           1 u   19   64  377   37.887  
>>>> -16011.   0.122
>>>> Is there anything I can do to decrease the convergence time?
>>> Little or nothing!  NTPD can, and sometimes does, take ten hours to 
>>> reach "steady state".  It needs about thirty minutes to find a 
>>16 second excursions are not the result of ntpd convergence delays. 
>>They indicate something seriously broken.  Possibly something else is 
>>trying to set the time.
> Thanks for the pointer David.  I added the local hardware clock (
> to the ntp.conf and that nailed it down.  Now my convergence is under a
> minute!

One problem in the sequence you showed was that your oncore gps started
off 1 second out from "true time". It then seems to have corrected
itself. This may well have confused ntpd into
starting a huge drift correction, which then played itself out. I do not
understand why the 128ms rule did not get applied. 

Certainly having a refclock be inconsistant to the 1 second level would
not do good things for ntpd.

More information about the questions mailing list