[ntp:questions] Reference Clock / shared memory driver

Unruh unruh-spam at physics.ubc.ca
Thu Jan 29 06:39:47 UTC 2009


stuart clark <stuart.clark at baesystems.com> writes:

>On Jan 29, 2:51=A0pm, Unruh <unruh-s... at physics.ubc.ca> wrote:
>> "Richard B. Gilbert" <rgilber... at comcast.net> writes:
>>
>> ...
>>
>> >Also note that ntpd needs at least thirty minutes after a cold start to
>>
>> More like 10 hours if it is to get down to usec accuracy. A few hours for
>> msec accuracy.
>>
>> >beat the clock into submission! =A0Depending on how cold the cold start
>> >was it may take longer; e.g. if you've been storing your computer in an
>> >unheated garage with the temperatures below -20 degrees F, ntpd might
>>
>> And that is with a warm clock.

>Sorry i dont understand.

>I if use the server 192.168.1.10 with a clinet clock one hour in front
>of the server, then after less than two minutes the client clock is
>set to the server clock.... BUT

Of course. If you are using -g an hour will be stepped. An hour is a silly
test I am afraid. Make it 40milliseconds.


>if i use the 127.127.28.2 with a client clock one hour in front it
>doesnt seem to update within the same "less than two minutes". maybe i
>have to wait longer???

Without the -g it will never do anything. And the shm PPS driver must get
to withing 128ms before it works at all.


>If so then is there something "special" about how the client ntpd uses
>the shared memory that is different to the way it uses an ip address
>ie the 192.168.1.10???

>cheers




More information about the questions mailing list