[ntp:questions] Re: Box not synchronizing

mooron at mooron.com mooron at mooron.com
Sat Jun 10 20:13:30 UTC 2006


John Pettitt wrote:
> Brian Utterback wrote:
> >
> > So, there you have it. I hope that helps you to understand how
> > a longer poll interval can lead to a more accurate clock, and
> > why it is a good idea to let NTP decide what the best poll interval
> > actually is.
> >
> >
>
> Great explanation - it also explains why sometime you do need to mess with minpoll and maxpoll.  If
> you have a machine that is prone to rapid changes in it's drift rate (typically due to thermal
> issues) when ntp moves to a longer poll it doesn't notice fast enough that the clock is drifting
> faster or slower that it was before.  This is very visible with PC based stratum 1 clocks where
> sporadic system loads (causing heat in the cpu) can push the clock off by hundreds of microseconds
> before ntp catches it and brings it back.  Setting minpoll and maxpoll to 4 for the refclock source
> can significantly improve the systems ability to remain relatively stable through these thermal events.

Yes, it sometimes takes NTP too long to figure out the clock is
drifting.
I've seen computer clocks off by as much as 60ms with NTP loafing
along with a polling interval of 1024 seconds.  NTP eventually reduces
the polling interval and brings the clock back in only to increase the
interal again and do the same thing over and over again. Locking the
interval at 64 seconds cures the problem.

- Mooron




More information about the questions mailing list