On 2011-07-06, Chris Albertson <[email protected]> wrote: > On Wed, Jul 6, 2011 at 8:41 AM, <[email protected]> wrote: >> Thank you for your answer, but can this limit be specified in the config >> file or command line, please? > > I found it. > http://www.eecis.udel.edu/~mills/ntp/html/clock.html > > Looks like I was wrong. From my read of the above 2 minutes is under > the "panic" limit and NTP should attempt to correct the offset. It > will "step" the clock in jumps of 125 milliseconds. So it would > take 1,000 steps to correct a 2 minute offset. So take a look every > few hours and you should see it slowly correct itself. As I wrote > above 2 minutes is a large and unrealistic test, try two seconds. > Two seconds is close to the error you might see in a network outage.
Unless the network outage lasts for weeks, no you would not expect to see 2 sec errors due to network outages. ntp should discipline the rate to a few PPM, so it would lose 2 sec in about 1000000 sec, which is about 12 days. > > Read the link above. It's explained there > > > The problem, or the reason NTP can't close a large offset is that NTP > tries to control the RATE of the clock. How can it close a 2 minute > offset unless it runs the clock faster than it should? This is a > fundamental problem because there is no "right" answer. The only > option is to correct one error (the offset) by creating another (A > clock that runs faster than one second per second). > > _______________________________________________ questions mailing list [email protected] http://lists.ntp.org/listinfo/questions
