[ntp:questions] ntpd: more than 60 s behind

David Woolley david at ex.djwhome.demon.invalid
Wed Apr 20 07:33:25 UTC 2011

asklucas wrote:

>    .ATOM.           1 u  781 1024  377   21.873  56147.1 3375.72
>  ptbtime1.ptb.de .PTB.            1 u  747 1024  377   43.859  52929.2 3343.98
>    .ATOM.           1 u  764 1024  377   21.023  52838.6 3394.36
>  ptbtime2.ptb.de .PTB.            1 u  768 1024  377   43.901  56232.9 3394.40
>  nav.metrologie. .ATOM.           1 u  749 1024  377   20.462  52937.8 3346.61
>  ptbtime3.ptb.de .PTB.            1 u  785 1024  377   43.870  56161.2 3384.54
>  ntp.nsm.pl   2 u  759 1024  377   47.775  52883.6 3359.28

Your offsets are all over the place.  It looks like your motherboard 
clock is beyond salvation, or you are trying to run ntpd on a virtual 
machine (whether this is a very silly thing or the best way of achieving 
some discipline on one is very controversial at the moment).

In the VM case, please read the archives (e.g. groups.google.com).

Ultimately, I think that ntpq rv will show that they are all being 
rejected because the jitter exceeds the maximum allowed error band.

> # your local system clock, should be used as a backup

This is bad advice.  It may be used, but only in certain unusual cases. 
  You should remove all references unless and until all three of the 
following are true:

1) you have solved the immediate problem;
2) you are serving time from the machine;
3) you have identified a specific problem that will benefit from its use.

The local system clock is not actually used as a backup, as it is always 
used as the time on the system.  What this does is to pretend that the 
machine is synchronised, for the benefit of downstream systems, when it 
isn't really.

One final point. Please do not use stratum 1 servers for a Windows based 
time server, and certainly never for a Windows based client.  Also don't 
do so for any VM based server.  The time keeping quality of such 
machines doesn't justify putting a load onto those servers.

More information about the questions mailing list