[ntp:questions] Ginormous offset and slow convergance
stenn at ntp.org
Wed Nov 30 21:28:15 UTC 2011
> I think the effect of getting rid of the drift file depends on the
> value stored in the file! If the value is reasonably close to
> correct, I think it's helpful. If you are restarting because you have
> been without power for the last three hours, the drift file is almost
> certainly useless!
Possibly, and in perhaps many cases probably, but almost certainly?
I suspect that if ntpd is not running on a machine that is expected to
be "on", then the odds are that the machine is hung or wedged, and a
small % of the time the problem is power.
It would be useful to have real data about the time it takes to get ntpd
sync'd and stable from a variety of startup cases where we compare
"leave the driftfile there" and "remove the driftfile" across this range
of startup cases. The next step would be for folks to decide what % of
time each of those startup times are likely to occur.
More information about the questions