[ntp:questions] Re: ntpstat versus ntpq -p

David Woolley david at djwhome.demon.co.uk
Sat Jun 11 08:14:38 UTC 2005

In article <T1118261529 at djwhome.demon.co.uk>,
david at djwhome.demon.co.uk (David Woolley) wrote:
> In article <1118225244.130834.218030 at g44g2000cwa.googlegroups.com>,
> Davide Bolcioni <db-lto at 3di.it> wrote:

> > *ntp1.test.xxx     2 u   26  512    77  81.371 885.281 265.723
> > +ntp2.other.xx     2 u   25  512    73 232.661 483.476 285.616

> The error bands on these two don't overlap so one, or both, of them has to
> be bad.  I'm not sure why either of them are shown as selected, especially

I wasn't strictly correct there, in that I ignored the root delay
and root dispersion, which extend the error bands.  However, even if
you assume that the local hop delays are zero in one direction for
one and zero for the other for the other, you still end up with the
upstream servers differing by 157ms, and if one assumes that the 81
ms round trip is near minimal, 237ms.  237ms would make at least one
of them outside the 128ms step out limit!  Even 157ms means one must
be more than half that limit away from true time.

They might not be about to step, because they may have very assymmetric
delays themselves, but at least one of them is certainly too far from
true time to be safe operating as a stratum 2 server!

More information about the questions mailing list