[ntp:questions] ntpd: time reset problem

Richard B. Gilbert rgilbert88 at comcast.net
Mon Sep 14 20:57:03 UTC 2009


David Lord wrote:
> Steve Kostecke wrote:
>> On 2009-09-14, David Lord <snews at lordynet.org> wrote:
>>
>>> Frank Elsner wrote:
>>>
>>> OT you also have rather a lot of sources specified and if just for a
>>> single box it seems a bit excessive and could be trimmed to four or
>>> five (I'd also guess some of those eight get same refid).
>>
>> Red Herring.
> 
> No, you missed the OT.
> 
>>
>>> I've just been trying mobile broadband on a high latency connection
>>> and it was too much for ntpd at 1-2 sec average and 150ms best to 10
>>> sec/timeout at worst :-)
>>>
>>> Then this weekend after reconfigured to use "burst" in server lines
>>
>> "burst" is an 8x multiplier that is applied to each poll.
>>
>> The use of "burst" against other peoples' time servers is generally
>> considered to be abuse unless you have received permission to do so.
> 
> Can you then suggest how it's at all possible for ntpd to be
> of any use on a mobile connection with such high latency?
> 
> Minpoll was also set to 8 which doesn't exactly compensate
> but does mitigate to some extent.
> 
> Probably from that quick test where time was set to within
> 20 ms rather than still 2 sec out after 30 min the poll
> interval could have been reduced further, however I'm pretty
> sure using other than "burst" would not have synced the time
> before system had to be shutdown.
> 
> I just need to get something working that's significantly
> better than wristwatch time.

NTPD was designed for the long haul.  It does not work very well on 
equipment that is not operated 24 hours per day.  If you want to power 
up at 8:30 AM and power down at 5:00 PM, NTPD cannot give you the 
accuracy that it is capable of when operated as designed.

When it is cold started, NTPD needs up to ten hours to discipline your 
clock with the accuracy it is capable of.  Once synchronized, your clock 
should be correct until you power off.

Wrist watch time can be very good indeed.  Walmart offers a "radio 
controlled" watch that synchronizes with the VLF timing signal provided 
by NIST on radio station WWVB for $40 or $50 US

You would do better to specify an error less than X milliseconds or X 
microseconds.

If you want to spend some money you can probably discipline your clock 
within an envelope of fifty or one hundred microseconds!




More information about the questions mailing list