[ntp:questions] NTP.conf using Dave Hart code.

Dave Hart davehart at gmail.com
Wed Feb 10 08:56:44 UTC 2010


On Feb 10, 2:39 UTC, David Lord <sn... at lordynet.org> wrote:
> Target was given above as +/- 10ms so just using
> "ntpd -q" will get there but here that seems to take
> between 5-8 minutes. Then after starting ntpd it's
> about another 4 minutes before ntpd is giving out
> time. As you say, it takes a few hours to reach
> offsets in low us. If an ntp server is configured
> with iburst, the startup time can be even faster.

ntpd syncs to my GPS+PPS much more quickly than that.  I'm guessing
you're using a substantially older version of ntpd than 4.2.6 and
probably also using the default minpoll 6 with your refclock drivers?

> These are from a restart yesterday morning:
>
> 09:54:40 clock PPS(0) 'clk_noreply' # ntpd started
> 09:58:57 synchronized to GPS_NMEA(0), stratum 0
> 10:00:18 synchronized to PPS(0), stratum 0

This is not directly comparable because it's from a warm restart of
ntpd where no step was needed, but in under 20 seconds it was humming:

10 Feb 05:05:40 ntpd[5988]: 0.0.0.0 c016 06 restart
10 Feb 05:05:40 ntpd[5988]: GPS_NMEA(1) 8024 84 reachable
10 Feb 05:05:40 ntpd[5988]: GPS_NMEA(1) 963a 8a sys_peer
10 Feb 05:05:40 ntpd[5988]: 0.0.0.0 c415 05 clock_sync
10 Feb 05:05:41 ntpd[5988]: PPS(1) 802b 8b clock_event clk_no_reply
10 Feb 05:05:57 ntpd[5988]: PPS(1) 8034 84 reachable
10 Feb 05:05:57 ntpd[5988]: PPS(1) 974a 8a sys_peer

The 16s vs. 64s poll intervals only explains part of the difference.
Another part is newer ntpd jumps on a working PPS signal as soon as it
is reachable, rather than after 4 polling intervals.

Cheers,
Dave Hart




More information about the questions mailing list