[ntp:questions] Mixed maxpoll values for mixed LAN/Internet servers - sensible?

David J Taylor david-taylor at blueyonder.neither-this-bit.nor-this.co.uk
Fri Feb 27 16:55:35 UTC 2009

Dave Hart wrote:
> On Feb 26, 11:27 pm, "David J Taylor"
> <david-tay... at blueyonder.neither- this-bit.nor-this.co.uk> wrote:
>> Well, by observation, /all/ the servers are polled at 64s intervals,
>> and not just the lan-gps-pps server.
> Since your clients are naturally preferring the stratum 1 on the same
> LAN, that server's poll interval is clamping your overall poll
> interval down with it.  In the similar configuration of a refclock
> plus internet servers, if you set minpoll 4 on a reference clock (and
> no maxpoll) the similarly unselected internet servers are allowed to
> progress to higher poll intervals.  It may not be possible to achieve
> the same without a reference clock.

Thats doesn't seem to happen with my FreeBSD system, if I am understanding 
you correctly.  Admittedly I have ntpd 4.2.0-a, but all servers have a 64s 
polls even though there is no minpoll or maxpoll specified.  My 
understanding was that having the physical GPS/PPS connection to that 
system forced this.

> My suggestion is to pick one ntpd on your LAN to sacrafice the best
> LAN timekeeping performance on while providing internet sources to
> your LAN in case of problems with the FreeBSD refclock box.

> Cheers,
> Dave Hart


Thanks for your suggestions - which are certainly worthy of consideration. 
However, apart from the Vista PC, I consider the timekeeping on my 
existing Windows PCs adequate for my needs - I'm not after sub-millisecond 
accuracy.  And of course, if I move to using one of the Windows PCs 
instead of the FreeBSD box as my stratum 1 server, I'll need to consider 
the configuration once again.

The clients certainly lock to the local stratum 1 clock provided by the 
FreeBSD server, and it would be nice if the other servers listed in those 
clients could be automatically demoted to a "check every so often just for 
sanity" mode.


More information about the questions mailing list