[ntp:questions] Re: tinker step 0 (always slew) and kernel time discipline

David Woolley david at djwhome.demon.co.uk
Sat Sep 23 08:29:35 UTC 2006


In article <RoSdnX8iZri9yonYnZ2dnUVZ_qWdnZ2d at comcast.com>,
Richard B. Gilbert <rgilbert88 at comcast.net> wrote:

> This just what happened.  Both servers seem to have been serving their 
> unsynchronized local clocks.  The servers diverged and the client had no 

>From what he said, I think that they *were* configured with a real server
as well, but that, at least one, real server was given a wrong address,
and maybe the stratum of the local clock reference wasn't set to 10 or so.

One of the points that I keep making is that configuring a local clock
is something that should only be done after considerable thought,
because, amongst other things, it prevents downstream ntpd's detecting that
the server in question is hopelessly unsynchronised.  Unfortunately,
most distributors seem to be enamoured with the idea and configure it 
in by default.

If the servers in question had not had a local clock configured, and had
a bad real reference configured, they would either never claimed to be
valid sources, or would have been detected as such within about a day of
losing their good server.  If time is important to the application, it
could have used the ntp_adjtime system call, or local equivalent, to 
detect whether there was a validly synchronised time, and alarmed at that
point.




More information about the questions mailing list