[ntp:questions] Consider the Undisciplined Local Clock synchronized at start-up

Steve Kostecke kostecke at ntp.isc.org
Fri Mar 30 19:58:04 UTC 2007

Eric <nospam-01 at jensenresearch.com> wrote:

> <devnull at localhost.com> wrote:
>>By definition, the undisciplined local clock is synchronized.
>>Therefore, one might expect that ntpd would consider it as such.
> Well, the local clock is "synchronized to itself" as the only point
> of reference.

The undisciplined local clock is not really synchronized to anything.
All the driver does, in large part, is return the current_time as the
poll_time and then set the necessary bits to allow ntpd to claim to be
synced to _something_.

> I think Steve was saying that it is normal for the NTPD
> process to take that period of time to synchronize with the local
> clock.

The undisciplined local clock is handled in the same way as all of the
other ref-clocks. ntpd has to poll the ref-clock 4 times before it is
possible to select that time source as the sys-peer. The first poll
occurs when ntpd starts up and the remainder of the polls take 3 * 64
seconds, or 192 seconds (a bit over 3 minutes).

You could use 'server minpoll 2' to reduce the default poll
interval to 16 seconds. So the initial sync time would be reduced to 3 *
16 seconds, or 48 seconds. On my ntp-4.2.2 test system this consistently
produced an intial "LocalCLK sync" of between 50 and 53 seconds.

BTW: The use of minpoll _may_ have some side effects.

>As for the other question, the verdict is still out.

Anyone may download the source from http://www.ntp.org/downloads.html or
http://ntp.isc.org/download and modify it in any way they please.

Steve Kostecke <kostecke at ntp.isc.org>
NTP Public Services Project - http://ntp.isc.org/

More information about the questions mailing list