[ntp:questions] Fifteen Second Leap

Andrew Hills ahills at ecs.umass.edu
Mon Sep 14 16:10:41 UTC 2009


Hello all,

I'm using NTP with a Motorola Oncore to keep a data logging system
synchronized. In general, it works great--peerstats reports offsets of
under 5us, which is much better than I need. However, between five and
twenty minutes of booting the system, the clock gets reset -15s, and I
can't figure out why. At this point, ntpq's peers reports that the
Oncore is being used as its source (with a reach of 377), and the
Oncore itself reports that it has been tracking satellites. The ntpd
log shows the following (for example):

 2 Jun 20:40:55 ntpd[3010]: synchronized to GPS_ONCORE(0), stratum 0
 2 Jun 20:41:20 ntpd[3010]: time reset +24.870267 s
 2 Jun 20:41:20 ntpd[3010]: kernel time sync enabled 0001
 2 Jun 20:43:01 ntpd[3010]: synchronized to GPS_ONCORE(0), stratum 0
 2 Jun 20:56:14 ntpd[3010]: time reset -15.000072 s

What I have discovered is that UTC and GPS time are off by exactly
15s; specifically, GPS time is 15 seconds in the future compared to
UTC because of the absence of leap seconds. This makes me suspect
that, for some reason, NTP doesn't decide to use UTC until many
minutes later. Now, I don't particularly care which standard it uses
as long as it picks one. These fifteen second overlaps in my data are
really obnoxious.

So, is my suspicion about this behavior correct, and can I force NTP
to make the correction either immediately or not at all?

--Andrew Hills



More information about the questions mailing list