[ntp:questions] NTP with GPS and RTC

unruh unruh at invalid.ca
Sat Apr 27 16:41:45 UTC 2013

On 2013-04-27, Harlan Stenn <stenn at ntp.org> wrote:
> unruh writes:
>> On 2013-04-26, Harlan Stenn <stenn at ntp.org> wrote:
>> > unruh writes:
>> >> Define occasionally. Unfortuneately ntpd requires about 5-10 hours to
>> >> rediscipline a clock to ultimate accuracy when the connection comes up
>> >> again, so if the connection time is shorter than that, chrony (assuming
>> >> you run linux or some unix) is a better choice ( faster lock and
>> >> discipline time). It also allows you to "trim by hand". Ie, if you can
>> >> phone into the device, and find it is say 1 min off, you can tell chrony
>> >> to correct that offset by hand.
>> >
>> > And Bill's statement above also depends on your definition of "better".
>> >
>> > My understanding is chrony will do an excellent job of tracking the
>> > upstream source it is listening to.  That is not necessarily the same as
>> > maintaining better system time.
>> Sorry, what other definition do you have of "better system time"? All
>> anything can do is to use a source and assume that that sourc is a good
>> source. 
> I haven't looked in a while.  What does chrony do about selecting
> between multiple servers?  What about clockhopping?
> For some, "better" may mean "My clock frequency is stable and I am
> tracking the middle of the clique" as opposed to "I am tighty latched
> to the server I am currently listening to."

Not sure myself. I am always running with one server being the local gps
server which is what gets selected. I have not spent time looking at
chrony's handling of which external clock it tracks, or exactly what it
does to select the time that it tracks. So I have not seen clockhopping,
but that says not much. And as far as I know this has not received any
attention in the past 5-10 years. 


> H

More information about the questions mailing list