[ntp:questions] Bizzare half second disagreement between ntp hosts

Brian Utterback brian.utterback at sun.com
Wed Jan 2 15:18:54 UTC 2008

Unruh wrote:
> I have a very weird situation. I am running a GPS PPS (Garmin GPS18LVM)
> with a few machines as a backup/initialization. 
> Sudeenly for about half and hour, my GPS failed for some reason ( still do
> not know what was wrong since it had come back on air by the time I noticed
> something wrong). Every hour I run a ntpq -p just to check that my gps is
> on air. I got this report.
>      remote           refid      st t when poll reach   delay   offset jitter
> ============================================================================ ==
> xtick.usask.ca   .GPS.            1 u 1003 1024  377   44.954    0.213 480.149
> +sanrail.com      2 u  993 1024  377    1.486  -479.03 479.917
> +raptor.tera-byt    2 u  322 1024  377   17.295  -480.35 0.766
> *zeus.yocum.org    2 u  390 1024  377   70.415  -481.02 1.230
>  SHM(0)          .PPS.            0 l 1415   16    0    0.000   -0.002 0.001
> Now I believe the tick.usask.ca result, since all of the machines which use
> mine as a source suddenly noticed a .48 second jump when my GPS failed. But
> why in the world would three systems all suddenly be out by .48 sec? 
> Doing a peers on them, one has a GPS as its source, one a .WWVB. and one an
> .ACTS. Why should all three suddenly be out by half a second?

Why to you believe tick.usask.ca? Look at the jitter, it recently had
a jump of 480ms. It seems much more likely that the one server is off
than three independent servers.

The sudden jump is likely due to the way PPS is handled. Remember that
the offset if the clock is determined by the time between the PPS
firing and the nearest second tick of the system clock. The definition
of "nearest" is the one for which the offset is less than .5 seconds,
either in the future or the past. So, at some particular threshold of
phase offset, the clock offset determined by the PPS will switch sign,
and jump from .5 to -.5 or vice versa.

Brian utterback

More information about the questions mailing list