[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[LDM #TMK-479992]: Monotonic clock confusion

This archive contains answers to questions sent to Unidata support through mid-2025. Note that the archive is no longer being updated. We provide the archive for reference; many of the answers presented here remain technically correct, even if somewhat outdated. For the most up-to-date information on the use of NSF Unidata software and data services, please consult the Software Documentation first.

  • Subject: [LDM #TMK-479992]: Monotonic clock confusion
  • Date: Fri, 29 Aug 2008 09:01:04 -0600
Mike,

> > The installation instructions say  that the host machine should have a 
> > monotonic clock
> > accurate to within 1 ms. To achieve this, the instructions suggest using 
> > ntpd
> >-- which seems like a contradiction because ntpd itself sets the clock back.
> 
> ntpd will adjust the clock.  This can be either setting the clock back or 
> forward.

Just to clarify, when used as intended ntpd(8) adjusts the clock slew rather
than the clock time.  The slew is the first-order derivative with respect to
time of the clock's time: it's the rate at which clock-time passes.  Ideally,
this value should be one; in reality, however, all clocks have values that are
close to, but not exactly, one.  By adjusting the slew, ntpd(8) guarantees
that clock-time will be monotonic.

Having said that, we've seen cases where the slew anomally of a system clock
was so bad (e.g., 10 seconds per hour) that ntpd(8) had no choice but to 
reset the clock -- causing a 0th order discontinuity in time.

I agree with Tom.  Your setup will work.  You might, however, have to
adjust your clock-setting parameters (e.g., how often the clock is reset)
depending on the rate of reconnections.

Regards,
Steve Emmerson

Ticket Details
===================
Ticket ID: TMK-479992
Department: Support LDM
Priority: Normal
Status: Closed