Building a strat1/2 time server

Wolfgang S. Rupprecht wolfgang.rupprecht+gnus200702 at gmail.com
Tue Feb 27 18:56:53 UTC 2007


"Weiner, Michael" <weinerm at ccf.org> writes:
> What I am trying to do actually is build an authoritative or
> reference time server that we can use here for testing daylight
> saving time adjustments in our environment.

Using ntp to test future daylight savings time transitions seems like
an awfully round-about way to do it.  First of all, ntp tries really
hard (too hard if you ask me) to never set a bad time.  All your
clients under test would have to be tested by restarting ntpd with the
-g option every time you wanted to jump the time back to before the
transition.  If you are going to go to all that trouble, you might as
well not run ntpd and just type "date <time-and-date>" and reset the
time by hand.

The other problem is that the stock Linux kernel doesn't talk to
reference clocks.  It needs some patches to do the PPS (Pulse Per
Second) protocol that reference clocks use to indicate the precise top
of the second.  

(As someone else mentioned, you can still use the phony "local"
refclock to have your server pretend it still has a valid time.  In
that case just set the time on your server to be the time under test,
restart ntpd and then restart the ntpd's on the clients under test
remembering to use the "-g" flag.)

Q: anyone know why the PPS patches never seem to make it into the real
kernel sources?

-wolfgang
-- 
Wolfgang S. Rupprecht                http://www.wsrcc.com/wolfgang/




More information about the fedora-list mailing list