[ntp:questions] Transparent NTP in-node forwarding

Steve Kostecke kostecke at ntp.isc.org
Fri May 27 01:54:16 UTC 2005

On 2005-05-27, Danny Mayer <mayer at gis.net> wrote:

> At 04:44 AM 5/26/2005, Jiri Hlusi wrote:
>>NTPD seems to bind to all available IP interfaces at its startup,
>>and there is no chance to change this later. Hence, in case that
>>the Virtual IP is introduced *after* local ntpd is started (due to
>>whatever reason), the only way to make ntpd to use it is to restart
>>the daemon as such. That's something I wouldn't like to do .....
> See bug #51 in bugzilla. This won't be fixed anytime soon. It doesn't
> take much to restart ntpd so why not do it for your situation?

If your ntpd has been able to create a drift file _and_ you are using
iburst on your server lines, an ntpd restart takes ~15 seconds to
achieve sync.

What Jiri is talking about sounds alot like the ntpd vs dhcp issue. The
best solution is to run your local master ntpd on a system behind the
system with the dynamic interface. This insulates your local master ntpd
from the IP address changes on that dynamic interface and, if you're
using NAT without port-forwarding, prevents outsiders from contacting
your local master ntpd.

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

More information about the questions mailing list