Folkert van Heusden schrieb: >>> It seems the Dutch NMi organisation (which is the time reference for the >>> Netherlands) has an NTP-service as well. Now I tried retrieving the time >>> with ntpdate (just to see if it was reachable) but ntpdate refuses it. >>> Using the regular ntp daemon works fine. Could someone enlighten me why >>> it is refused? I mean: "has gone too long without sync" seems a little >>> odd as it is supposed to be connected to real atomic clocks. > ... >>> reference time: cd0c473b.73087696 Mon, Jan 5 2009 9:45:47.449 > ... >> A wireshark capture shows that it sends a bogu reftime ... >> Reference Clock Update Time: Jan 5, 2009 08:45:47,4493 UTC >> Originate Time Stamp: Jan 8, 2009 10:15:39,0507 UTC >> Receive Time Stamp: Jan 8, 2009 10:15:39,0375 UTC >> Transmit Time Stamp: Jan 8, 2009 10:15:39,0376 UTC >> The Root dispersion does not look too healthy, too... >> Root Dispersion: 3,9689 sec >> This really looks like they should have a look at their NTP server and its >> IRIG >> source. Why this server is accepted by ntpd is a miracle for me. Only chance >> I >> would see is if you have no other sources configured. > > Got this morning an e-mail from NMi and what they say is that they only > occasionally connect their ntp server to a source that says what time it > is. They had it connected to that irig-b because of the leapsecond and > disconnected it at January 5. The rest of the time they only have it > connected to the PPS source.
But it seems that they are using something else than NTP to synchronize the time on their NTP server to this PPS signal, otherwise the RefID should say "PPS" but should be rejected because the the ToD source for this PPS reference is not synchronized anymore. > That root dispersion, does that mean the time of that clock is almost 4 > seconds behind the real time (the time of the IRIG)? From RFC1305: "Root Dispersion (sys.rootdispersion, peer.rootdispersion, pkt.rootdispersion): This is a signed fixed-point number indicating the maximum error relative to the primary reference source at the root of the synchronization subnet, in seconds." In the NTPv4 draft you will find a (similar) definition: Root dispersion indicates the maximum error, that does not necessarily mean that this is the current error. As mentioned by other people, I would recommend to use the local clock if they use some none-ntp means to steer the clock of their NTP servers with their undoubtedly excellent PPS signal. But it is mandatory that they introduce some means to be able to stop claiming that they are stratum 1 when the PPS input fails. The preferred solution IMHO would be to permanently connect their IRIG source to the server and probably use the ATOM driver instead of mixing NTP and non-NTP references. Cheers, Heiko > Folkert van Heusden > _______________________________________________ questions mailing list [email protected] https://lists.ntp.org/mailman/listinfo/questions
