[ntp:questions] Re: Two clocks?
Richard B. Gilbert
rgilbert88 at comcast.net
Tue Nov 16 06:18:07 UTC 2004
Harlan Stenn wrote:
>I'd be inclined to connect them to different machines (and have them peer
>with each other) so in case one of them went down you'd have the other
>up and still serving time to the rest of your network.
That was my first thought. I raised the question after actually trying
it. I started with machine "sunblok" configured with the Motorola M12+T
GPS receiver and all the servers, good and not so good, in the "greater
New Jersey" area. That gives me five networks servers plus the
reference clock and peering with the second machine, "sunburn".
Today I configured Sunburn with the Traconex and peered with sunblok. I
didn't want to tap any of my network servers a second time since they
are all busy and some appear to be staggering. Sunburn is running ntp
4.2.0a at 1.122. I saw a situation earlier today where both the refclock
and the peer were "insane":
remote refid st t when poll reach delay offset
xWWV_PST(0) .WWV. 0 l 52 64 377 0.000 67.135
xsunblok .GPS. 1 u 947 1024 377 0.518 -59.704
As someone here remarked recently, two is the worst possible number of
servers! With my "godlike" view of the matter, I can safely say that
the WWV receiver is probably having the problems and that sunblok should
be believed. I can say that because I know from several months of
experience that when sunblok is synched with GPS it generally has an
offset of less than one millisecond. ntpd on sunburn has no way to
figure that out.
More information about the questions