[ntp:questions] Linux/Windows: Different synchronization behaviour
maarten at kittensandcats.net
Tue Oct 31 10:00:56 UTC 2006
> "Geir G" <geir.guldstein at start.no> wrote in message
news:1162279048.696498.218770 at k70g2000cwa.googlegroups.com...
> We are running NTP in an isolated network. We use 2 HOPF stratum 1
> NTP servers with GPS ...
> Below I attach ntpq output and the ntp.conf configuration file. The
> Linux ntpd is happy and uses the working HOPF server as its system
> peer, while the Windows ntpd considers both HOPF servers to be
Well, that's what you get for using two, I'm afraid.
> ntpq output from Linux:
> ntpq> pe
> remote refid st t when poll reach delay offset
> *nccgps02 .hopf. 1 u 30 64 77 7.812 -370.45
> +nccgps01 LOCAL(0) 11 u 35 64 77 7.812 629.355
They're a second apart! No wonder the software gets confused.
IIAMN, the stratum is merely informative and there is no quantifiable
reason to prefer a stratum-1 server over a stratum-11 by itself. All
sorts of error margins might be reported as zero on the disconnected
clock, resulting in it looking perversely attractive rather than the
bad choice it really probably is. (The problem is in the 'probably'.
It _might_ be synchronised very well by external means. NTP doesn't know.)
The best solution would probably be to turn off the local-clock fallback
in the Hopf units.
More information about the questions