[ntp:questions] Re: Ref clocks and polling intervals

David L. Mills mills at udel.edu
Wed Nov 24 22:18:49 UTC 2004


Chris,

The reference clock poll interval is not intended to increase unless you 
tell it to. That might be a bug; I'll look into it. However, just about 
all the machines around here loaf at 1024 s without swinging more than a 
couple of milliseconds. If yours are much greater than that, including 
the reference clock, something else might be going on. Set the maxpoll 
for the reference clock to 6 and leave the others alone. You should see 
the same behavior as I.

Dave

Chris Hastie wrote:
> About a year ago I raised an issue I was having whereby as soon as I
> connected a ref clock the poll interval for all other sources dropped to
> 64s and stayed there.
> 
> A message from David Mills that I missed was recently brought to my
> attention and as a result I upgraded to a dev version announcing itself
> as 4.2.0 at 1.1179-r Wed Nov 17 16:04:21 GMT 2004 (1).
> 
> This has led to poll intervals rapidly reaching 1024s. But as David
> points out in the message, it stiffens the PLL frequency and
> consequently I'm seeing a marked diurnal oscillation of offset - the
> machine is no where near air conditioned, in fact it's in a room that
> the woodburner's chimney goes up through :( See <http://ntp.raggedstaff.
> net/offsets2.php>
> 
> 10 or 11 hours ago I set maxpoll to 8 on the reference clock in an
> attempt to find a compromise. It's early, but it looks as if this may
> have calmed down the daily swings a bit. But of course it's also
> increased the frequency of polling to other sources.
> 
> Can anyone give me some hints as to what effects the following might
> have:
> 
> 1)  Reduce maxpoll for the reference clock to 6, but explicitly set it
>     to 10 for all other sources.
> 
> 2)  Reduce maxpoll for the reference clock to 6 but use a tinker maxpoll
>     10 command
> 
> What I'm trying to achieve is a reduction in the daily swings whilst
> still being a better neighbour than with a 64s poll interval to
> everything.
> 




More information about the questions mailing list