[ntp:questions] Re: ntp servers reporting leap second erroneously?
martin.burnicki at meinberg.de
Mon Oct 17 15:33:57 UTC 2005
Danny Mayer wrote:
> I do have reservations about stepping the clock *BACK* by one second.
So do I.
However, the way it works right now is a big improvement compared to the way
it has just worked a few days ago, when the system time was stepped back
*and* the clock frequency was totally messed up until a second time step a
few minutes was required to correct that.
> would rather see that the clock increments more slowly until time
> catches up with it since time is always supposed to be a monotonically
> increasing function. This can also impact logging when tracking done
> suspicious events, database timestamps, etc.
I'd also like to see in the syslog when ntpd handles a leap second, either
by notifying the kernel, or by slewing the time.
Please see also the logs for w32time which I've posted in a reply to Michael
Wouters (Arthur.Porkchop at exemail.com.au) on Oct 14, where w32time slews the
system time extremely to catch up quickly.
So a few seconds after the leap second event the time is synchronized again,
it has been handled in a short time, and the system clock has increased
monotonically without discontinuity.
In the special case of a leap second I would also allow for a slewing rate
which exceeds the normal maximum rate in order to follow up as quickly as
More information about the questions