f



Question on how the NTP daemon syncs the system time with an NTP Server(s)?

Does anyone know whether the NTP daemon syncs the system time with the * marked NTP server (found with the "ntpq -p" command), or does it perform calculations (like an average) using all NTP servers marked with an * and +?

I've seen reference in the NTP.ORG debugging pages, indicating that the NTP daemon does a calculation using multiple servers.  And I've seen other references stating that the daemon syncs to a single NTP Server.

Thanks!

jc
0
jchock2000
11/10/2016 10:03:46 AM
comp.protocols.time.ntp 4895 articles. 2 followers. Post Follow

4 Replies
194 Views

Similar Articles

[PageSpeed] 5

jchock2000@gmail.com wrote:
> Does anyone know whether the NTP daemon syncs the system time with the * marked NTP server (found with the "ntpq -p" command), or does it perform calculations (like an average) using all NTP servers marked with an * and +?
> 
> I've seen reference in the NTP.ORG debugging pages, indicating that the NTP daemon does a calculation using multiple servers.  And I've seen other references stating that the daemon syncs to a single NTP Server.

AFAIK, older versions of (x)ntpd used the system peer ('*') exclusively
to adjust the system time.

Current versions (v4.x) indeed compute the adjustment based on the
weighted polling results from of the system peer ('*' o r 'o') and the
candidates ('+').

This is particularly *bad* if you have a GPS/PPS time source which is
very accurate, and additional NTP servers on the network have also been
configured as time source.

The weighted polling results from the NTP server(s) have some jitter and
thus the resulting weighted time offset is less accurate, resulting in a
less accurate adjustment of the system time.

Martin
-- 
Martin Burnicki

Meinberg Funkuhren
Bad Pyrmont
Germany
0
Martin
11/10/2016 2:36:07 PM
On 10/11/2016 14:36, Martin Burnicki wrote:
[]
> AFAIK, older versions of (x)ntpd used the system peer ('*') exclusively
> to adjust the system time.
>
> Current versions (v4.x) indeed compute the adjustment based on the
> weighted polling results from of the system peer ('*' o r 'o') and the
> candidates ('+').
>
> This is particularly *bad* if you have a GPS/PPS time source which is
> very accurate, and additional NTP servers on the network have also been
> configured as time source.
>
> The weighted polling results from the NTP server(s) have some jitter and
> thus the resulting weighted time offset is less accurate, resulting in a
> less accurate adjustment of the system time.
>
> Martin

Martin,

Is there no option to revert to the earlier behaviour?

-- 
Cheers,
David
Web: http://www.satsignal.eu
0
David
11/11/2016 6:41:19 AM
On 2016-11-11, David Taylor <david-taylor@blueyonder.co.uk.invalid> wrote:
> On 10/11/2016 14:36, Martin Burnicki wrote:
> []
>> AFAIK, older versions of (x)ntpd used the system peer ('*') exclusively
>> to adjust the system time.
>>
>> Current versions (v4.x) indeed compute the adjustment based on the
>> weighted polling results from of the system peer ('*' o r 'o') and the
>> candidates ('+').

> Is there no option to revert to the earlier behaviour?

I have not tried it, but I think setting tos minclock to 1 should do
that.

-- 
Miroslav Lichvar
0
Miroslav
11/11/2016 9:08:03 AM
Miroslav Lichvar wrote:
> On 2016-11-11, David Taylor <david-taylor@blueyonder.co.uk.invalid> wrote:
>> On 10/11/2016 14:36, Martin Burnicki wrote:
>> []
>>> AFAIK, older versions of (x)ntpd used the system peer ('*') exclusively
>>> to adjust the system time.
>>>
>>> Current versions (v4.x) indeed compute the adjustment based on the
>>> weighted polling results from of the system peer ('*' o r 'o') and the
>>> candidates ('+').
> 
>> Is there no option to revert to the earlier behaviour?
> 
> I have not tried it, but I think setting tos minclock to 1 should do
> that.

That's interesting, I didn't know this before. It should be really
helpful if that worked.

Thanks,

Martin
-- 
Martin Burnicki

Meinberg Funkuhren
Bad Pyrmont
Germany
0
Martin
11/11/2016 10:17:19 AM
Reply: