[ntp:questions] Re: Oncore leap second bug at 11/29/2003 62:28:15

Tapio Sokura oh2kku at iki.fi
Sat Nov 29 00:37:36 UTC 2003


Tom Van Baak wrote:
> Sure enough, the Oncore 256 week leap second
> bug showed up as predicted by Motorola. It was
> a little different than we thought; this is how an
> Oncore VP GPS receiver reported time yesterday:
> 
> 11/27/2003 23:59:58
> 11/27/2003 23:59:59
> 11/29/2003 62:28:15
> 11/28/2003 00:00:00
> 11/28/2003 00:00:01
> 
> If any of you saw NTP anomalies I'd be interested
> to hear about them. Presumably the invalid hour
> saved the day (the time stamp would be rejected).

I'm running an Oncore VP (6 channels) with version 8.4 firmware and ntpd 
4.1.1a at 1.791 in Linux with nanokernel (2.2.19). My ntpd clockstats file 
had these entries in it at the change:

(last line of 2003-11-27)
52970 86399.100 127.127.30.1 3278966398.999990798 2003 331 23 59 59 58 
rstat   08 dop  0.0 nsat  7,2 traim 2 sigma 65535 neg-sawtooth -13 sat 
5000051470

(first line of 2003-11-28)
52971 1.100 127.127.30.1 3278966400.999997669 2003 332  0  0  0  0 rstat 
   08 dop  0.0 nsat  7,2 traim 2 sigma 65535 neg-sawtooth -15 sat 5000052100

Compared to normal it looks like one timestamp was completely left out 
of the log (the zero second stamp). loopstats shows a jump in offset 
just after the UTC day change from the regular +10 us to about -2 ms for 
a while:

52970 86324.106 0.000014667 -264.099197 0.000000821 0.005368 4
52970 86340.099 0.000012934 -264.099197 0.000001389 0.004648 4
52970 86356.100 0.000010806 -264.099197 0.000003048 0.004026 4
52970 86373.100 0.000012878 -264.099197 0.000001656 0.003486 4
52970 86390.107 0.000012850 -264.099197 0.000003494 0.003019 4

52971 7.099 -0.000662972 -264.099197 0.000002537 0.002615 8
52971 36.310 -0.002153100 -264.108124 0.000001896 0.005005 8
52971 477.106 0.000026050 -264.149780 0.000003183 0.021274 4
52971 495.101 0.000027275 -264.149780 0.000000662 0.018424 4
52971 511.106 0.000026562 -264.149780 0.000001356 0.015956 4

I guess ntpd lost the ball for a while because there are no loopstats 
lines between seconds 36 and 477. This might also have something to do 
with the Oncore losing satellites; according to clockstats it was locked 
to zero satellites for about 200 seconds starting six seconds after the 
day change. Oncore losing the satellites every now and then is normal in 
my setup because the antenna is placed inside the house.

Nothing was output to the system log when this happened. I wasn't around 
when this event actually took place, just browsed through the logs 
afterwards.




More information about the questions mailing list