[ntp:questions] question regarding NTP configuration for clusters, and "cluster time" stability
unruh-spam at physics.ubc.ca
Sat Nov 21 18:01:52 UTC 2009
Steve Kostecke <kostecke at ntp.org> writes:
>On 2009-11-21, <Null at BlackList.Anitech-Systems.invalid> wrote:
>> The article I referenced seemed relevant to what they are attempting.
>> <http://lopsa.org/node/1480> When the external server(s) go away, the
>> internal servers take some number of minutes, but eventually elect a
>> leader and settle on a usable hierarchy. They may drift in total from
>> the absolute notion of NTP time, but they remain synchronized with
>> each other.
>What you really want is a decent quality PPS generator (e.g. Rubidium)
>which feeds an internal time server. That time server can become the
>internal master when the connection to the outside world goes away.
It you have an rubidium clock, you want it to be the master always, with the
outside net the backup and the internal orphan mode the ultimate fallback.
Of course a GPS is a lot cheaper than an rubidium clock (and probably more
accurate over the long term).
>Keep the orphan cloud as a back-up in case both the outside world and
>the internal master go away.
>Steve Kostecke <kostecke at ntp.org>
>NTP Public Services Project - http://support.ntp.org/
More information about the questions