[ntp:questions] Proposed NTP solution for a network

Unruh unruh-spam at physics.ubc.ca
Wed Mar 4 22:37:49 UTC 2009

Augustine <evandro at mailinator.com> writes:

>On Mar 3, 4:01=A0pm, "Richard B. Gilbert" <rgilber... at comcast.net>
>> NTP does this ONLY when the time is off by 128 milliseconds or more. =A0I=
>> NTP is that far off, other than at startup, something is badly wrong
>> somewhere!

>The problem is that I've seen this in the silly reaction of different
>NTP versions and servers to the recent leap second, causing my systems
>to be stepped by 1s back and forth (see bug #1110).

That is certainly an indiction that something is seriously wrong somewhere. 
ntp should NOT be reacting to leap seconds except at midnight UTC on June
30 or Dec 31. Now if a bunch of computers that  you are useing as a server
disregards the leap second, so that the majority of computers is not
applying the leap second, although yours did, then I could see that there
could be thrashing. Do not use those sources as your server. Note that ntp
and UTC has no memory. Ie, they do not know whether or not a leap second
has been applied, and have to trust that their servers have done so. 

More information about the questions mailing list