[ntp:questions] questions] Leapsecond on FreeBSD or Windows - no showstopper bugs, but ...

Dave Hart hart at ntp.org
Tue Jul 3 00:31:39 UTC 2012


On Sun, Jul 1, 2012 at 6:50 PM, David J Taylor
<david-taylor at blueyonder.co.uk> wrote:
> Dave Hart suggested I post the following event log information about leap
> second behaviour on various Windows stratum-1 and stratum-2 PCs here......
>
> ___________________________________________________________
> Stratum-1 server PC Alta (Win-7/64 + Sure GPS) (runs 24 x 7)
> Level    Date and Time    Source    Event ID    Task Category
> Warning    01/07/2012 01:24:33    NTP    2    None    clock would have gone
> backward 1 times, max 1000612.1 usec
> Information    01/07/2012 01:24:32    NTP    3    None    HZ 64.102 using 43
> msec timer 23.256 Hz 64 deep
> Information    01/07/2012 01:00:00    NTP    3    None    Leap second
> announcement disarmed
> Warning    28/06/2012 00:46:08    NTP    2    None    clock would have gone
> backward 1 times, max 32.5 usec
>
> ___________________________________________________________
> Stratum-1 server PC Stamsund (Win-7/32 + GPS 18x LVC) (runs 24 x 7)
> Level    Date and Time    Source    Event ID    Task Category
> Warning    01/07/2012 01:23:18    NTP    2    None    clock would have gone
> backward 1 times, max 722520.0 usec
> Information    01/07/2012 01:21:38    NTP    3    None    HZ 64.000 using 43
> msec timer 23.256 Hz 64 deep
> Information    01/07/2012 01:00:00    NTP    3    None    Leap second
> announcement disarmed
> Warning    29/06/2012 17:53:16    NTP    2    None    clock would have gone
> backward 1 times, max 10.9 usec
> Warning    29/06/2012 05:53:16    NTP    2    None    clock would have gone
> backward 1 times, max 10.4 usec

So there are at least two problems here.  First the insertion is not
happening, second when a 1s step occurs later, there's a "would have
gone backward" that shouldn't be reported.  I have an idea how to fix
that part.  For the first part, I'm leaning toward integrating the
Windows port's leap second insertion more with the POSIX daemon loop
code (while preserving the differing implementations).

> ___________________________________________________________
> Stratum-2 PC Torvik (Win-8/32) - booted especially for the event!
> Level    Date and Time    Source    Event ID    Task Category
> Information    01/07/2012 01:00:00    NTP    3    None    Inserting positive
> leap second.
> Information    30/06/2012 03:55:33    NTP    3    None    Detected positive
> leap second announcement for 2012-07-01 00:00:00 UTC
> Information    30/06/2012 03:55:31    NTP    3    None    peers refreshed
> Information    30/06/2012 03:55:31    NTP    3    None    Listen normally on
> 6 Loopback Pseudo-Interface 1 127.0.0.1:123
>
> ___________________________________________________________
> Stratum-2 PC Ystad (Win-7/32) (runs 24 x 7)
> Level    Date and Time    Source    Event ID    Task Category
> Information    01/07/2012 01:00:00    NTP    3    None    Inserting positive
> leap second.
> Information    30/06/2012 01:00:13    NTP    3    None    Detected positive
> leap second announcement for 2012-07-01 00:00:00 UTC
> Information    22/06/2012 16:09:36    NTP    3    None    peers refreshed

I saw different behavior on this laptop -- the insertion occurred, but
33 seconds late.  I can't rule out that having three thousand
"noselect" associations wasn't part of the reason, but I doubt it at
this point.

Thanks for the details,
Dave Hart


More information about the questions mailing list