On Mon, 23 Oct 2017, Rob Janssen wrote:
Bill Unruh wrote:
Ok but rather than "only a few hours" I would like to see "only a few
minutes".
But that would be totally rediculous. The offset of the local clock from
UTC
after even a few hours is still far far better than that from the network,
and
far better even than 20us. Remember what you want to know is how far the
local
clock is from UTC, not whether or not the local clock has not heard from
PPS
in the past few minutes.
You don't support my calculation that if the clock apparently wandered away
3400us
Again, no evidence of that 3400 us.
From the evidence that chrony has, pps does NOT wander that badly in 13 hrs.
Remember chrony constantly measures both the standard deviation in the offset
AND in the rate. So it has a good estimate of how far the offset will wander
in that time. And it is NOT 3400us. So you need to tell us how you measure
that 3400us.
after 13 hours, it would take about 5 minutes to wander 20us?
Not it would not. chrony has measured it, and it is not that much.
I would think it is a best-case calculation as it assumes a linear drift in
one direction.
I practice it will probably wobble, and take less than 5 minutes to wander
20us.
Please note we are talking MICROseconds here. Not MILLIseconds. I don't
think
many standard systems will remain within 20us for several hours if left
without sync.
(it would likely require some TCXO clock option)
Sure they could. If the temp is constant, as you claim, that is main cause of
changes in drift rate.
The Span indicated by sourcestats is 79 for the PPS source now, and 103m
for
the network sources.
Would that mean it drops the PPS after 79 seconds? That would be fine.
No. You really need to think through what you want and what the time on
your
server machine delivers. After all if the computer clock in your local
machine
was and exact track of UTC always to atto seconds, and you used the GPS
only
to make determine the intial offset determination then it would be silly to
throw away that source just because the pps had not been heard from.
We are not interested in "time that is likely a good estimation". We require
accurate time and if
I am sorry, but nothing will give you "accurate time" Not even GPS. What it
can give you is an estimate of the time and the accuracy of that estimate.
we do not have it, or do not have certainty about it, we need to shutdown our
application.
So we require some monitoring. Of course I can add monitoring of "sources"
or "sourcestats"
to the monitoring of "tracking" that we currently do, and alert when "Reach"
of the PPS
clock is zero. That is probably our quickest solution. However, I would
have expected this
error condition (missing PPS pulses) to be somehow reflected in the
"tracking" output.
Why?
Rob
--
To unsubscribe email chrony-users-requ...@chrony.tuxfamily.org
with "unsubscribe" in the subject.
For help email chrony-users-requ...@chrony.tuxfamily.org
with "help" in the subject.
Trouble? Email listmas...@chrony.tuxfamily.org.
