[ntp:questions] Wrong time after changing hardware
david at ex.djwhome.demon.co.uk.invalid
Sat Jul 21 08:52:16 UTC 2007
In article <ywn9hcnyzi5b.fsf at ntp1.isc.org>,
Harlan Stenn <stenn at ntp.isc.org> wrote:
> Not at all - ntpd can handle being off by at most 500ppm. If the old
> machine needed a drift value of, say, -300 and the new machine needed a
> value of 250 then ntpd will not be able to correct for this.
Yes it will. The new correction is less than 500ppm; the fact that the
change in correction exceeds 500ppm doesn't matter, as long as the actual
correction is less than 500ppm.
In fact, ntp won't ever compute the 550ppm; rather it will compute small
adjustments that build up, until it loses lock, at which point it will
calculate the correction from first principles.
The other questionable statement on this thread is that ntpd cares about
the technology used to implement the OS timers. Whilst it is true that
it does so on Windows (because Windows clock reading precision is too
low), I believe it is still the case that, on Unix-like systems, that
is the responsibility of the kernel.
More information about the questions