Re: [chrony-users] Discover when a time sync has been performed

[Ignacio Verona Ríos <ignacio.verona@xxxxxxxxxxx>]

2013/7/4 Bill Unruh <unruh@xxxxxxxxxxxxxx>

On Thu, 4 Jul 2013, Ignacio Verona Ríos wrote:

Yes, the GPS is pure NMEA, no PPS info is available :(

That should give you the time to about 50 ms. But  you say that the gps is not
always available? Where do you get the time from in that case?
Wrist watch? the modem? If the latter then you shouldstart up the modem before
starting up chrony, then use the initstepslew directive to get the time from
the modem.

Does the gps has contiuous power from the car battery or is it plugged into
the Pi for its power? Or is it a gps on your phone?

At boot time a first connection is done to let chrony get the time. I don't have any other reliable time source. GPS is attached to the Pi, as well as the modem. 

 

Situation is the following: In-car device, with GPS+3G modem, making some
tests/measurements and sending them to a remote server. So, I can not wait
for chrony to set itself, and that's why I ended up using makestep.

chrony can set itself pretty fast with the initstepslew directive. You could
make the timeout 1 sec if  you wished. At bootup the RPi will be out by years
anyway.

I'll play with the initstepslew directive to try to achieve a quick time convergence.

Thanks for your time!

 

2013/7/4 Bill Unruh <unruh@xxxxxxxxxxxxxx>

On Thu, 4 Jul 2013, Ignacio Verona Ríos wrote:

 Hi Bill,

yes, the rapspi is sometimes powered-off, specially now while I'm testing
the software. No process other than chrony is modifying the time, and
sometimes the board is powered-off for many hours (overnight), and so will
be the real behaviour of the final app.

Ah. So you want chrony to start up with the initstepslew  directive in
chrony.conf with a low time to allow it to set itself initially. You should
sometimes leave it on for a while so that the drift file can be properly
populated. But all of the ntp programs are designed to discipline a clock
over
a long period of time, not in a stop and go fashion. It is ok with chrony
if
it is only connected to the net sporadically (but for the initstepslew it
should be connected when it boots up).

I'm thinking about buying an RTC clock. They are cheap (15€) and maybe
connecting it to the GPIO ports help... but I'd like to stay with a pure
software source.

The problem is that none of the program can do anything if there is no time
source of any kind. They cannot pull time out the air.

You say you have  a gps attached. Does it have PPS? or is it just the nmea
sentences? Is it also switched on when the pi is switched on or is it on
permanantly? They also take a while to get a decent time.

Ie, telling us in some detail what your situation is might allow us to give
better advice.

2013/7/4 Bill Unruh <unruh@xxxxxxxxxxxxxx>

 On Thu, 4 Jul 2013, Ignacio Verona Ríos wrote:

 Thanks again! Yes, I'm using 1.24 and now realized the log is there:

Jul  3 20:39:37 m2m-probe1 chronyd[2077]: System clock was stepped by

 829.745 seconds

Jul  3 21:12:49 m2m-probe1 chronyd[2023]: System clock was stepped by

 3057.778 seconds

 That is looking seriously sick. It steps the clock, and 33 minutes

later it
steps it by 50 minutes?

 Jul  3 21:34:59 m2m-probe1 chronyd[2108]: System clock was stepped by

 21.122 seconds

Jul  3 22:46:03 m2m-probe1 chronyd[1983]: System clock was stepped by

 1698.671 seconds

Jul  4 09:17:25 m2m-probe1 chronyd[2047]: System clock was stepped by

 32109.587 seconds

 and here by 9 hours? Are constantly switching off and on you RPi?

 Jul  4 10:27:34 m2m-probe1 chronyd[1992]: System clock was stepped by

 36102.593 seconds

 and an hour later by 10 hours?

Is your program altering the clock itself?




 Jul  4 13:41:12 m2m-probe1 chronyd[2015]: System clock was stepped by

 3942.919 seconds

 Thank you very much!

2013/7/4 Miroslav Lichvar <mlichvar@xxxxxxxxxx>

 On Thu, Jul 04, 2013 at 04:32:04PM +0200, Ignacio Verona Ríos wrote:

 Hi Miroslav,

thanks for your answer. It partially solves my question, but I'm not
able
to see the entry in the tracking log when makestep is applied. Last
lines
of the file look like this:

 2013-07-04 13:05:17 193.145.56.7     3     82.620     46.004
-2.830e-03

2013-07-04 13:05:21 193.145.56.7     3     82.127     43.920
-2.164e-03

2013-07-04 13:05:25 193.145.56.7     3     81.555     42.269
-1.921e-03



 Should I expect some "Makestep OK" message? or it's a different
state?

Not in the tracking log, but in the system log there should be a
"System clock was stepped by x seconds" message.