[ntp:questions] GPS driver failure mode question

Steve Kostecke kostecke at ntp.org
Sun Jan 11 02:15:08 UTC 2009


On 2009-01-08, Charles Brown <charles.brown at sensis.com> wrote:

>    I have a processor with a serial GPS on a closed private network.  If 
> the GPS is not locked, or the GPS inputs are missing, I still want the 
> ntpd to serve *some* time, perhaps from the undisciplined local clock. 

ntpd _never_ serves time "from" the Undisciplined Local Clock. All that
the Undisciplined Local Clock driver does is allow ntpd to claim to be
synced in the absence of a real time source.

> I.e., the server can't just do nothing.  Does the GPS driver do that 
> automatically, or do I have to configure two drivers in the ntpd.conf? 
> Perhaps something like;
>
>          prefer peer 127.127.20.u
>          peer 127.127.1.u

That's all wrong. Use something like this:

# GPS
# see the type-20 driver page for the mode value
server 127.127.20.0 mode X minpoll 4 maxpoll 4

# Undisciplined Local Clock
server 127.127.1.0
fudge 127.127.1.0 stratum 10

>     Suppose I have a second server, with a higher precision oscillator, 
> on this same network. How would I configure that?  Perhaps;
>
>          prefer peer morePreciseServer   # should be best source
>          peer 127.127.20.u               # next is the GPS
>          peer 127.127.1.u                # if all else fails...

The best thing to do is let NTP choose the best time source. But keep in
mind that NTP needs at least 3 time sources to choose the one that is
the best. When NTP has only two time sources it has no way of deciding
which one is correct.

>What would the configuration on the other processor look like? Perhaps;
>
>          prefer peer 127.127.22.u  # always best
>          peer 127.127.1.u          # if GPS fails, still better for a
>                                    #   couple of days
>          peer lessPreciseServer    # GPS missing a long time,
>                                    #   lessPreciseServer might have GPS

I don't understand this configuration example.

Unless the "better" system is using an _extremely_ high quality
oscillator is will most likely not be materially "better" than the fist
system.

-- 
Steve Kostecke <kostecke at ntp.org>
NTP Public Services Project - http://support.ntp.org/




More information about the questions mailing list