Periodic spike in offset for pps

[Paul Theodoropoulos]

On 5/5/2020 11:08 AM, Doug Curtis via users wrote:
> I apologize if this has been answered previously.
>
> I've setup a few ntp servers on raspberry Pis using the ntpsec howto 
> with clockmaker.  Currently, I'm using a pi2, pi3 and pi4 with 2 of them 
> using the Adafruit gps hat and one of them using a "Gowoops GPS Module 
> <https://www.amazon.com/Gowoops-Module-Antenna-Arduino-Microcomputer/dp/B01MRNN3YZ/ref=cm_cr_arp_d_bdcrb_top?ie=UTF8>".  
> I setup ntpviz to run on a separate machine and I'm noticing a regular 
> offset spike for shm(1) in my graphs on all of them.
>
> I can't think of anything else to look for or try.
>
> Thanks,
>
> Doug
>

A few things to check, and questions.

Your temps seem quite consistent - are you using ntpheat or another 
application to try to smooth daily temperature variations? It would be 
unlikely they'd cause dramatic spikes though...

Only tangentially related  - since only two of the three are using the 
Adafruit gps hat - are you regularly (weekly, biweekly) downloading the 
satellite ephemeris data from mediatek?

The most peculiar characteristic is the 'long wave' rise of positive 
offset spikes followed later by negative offset spikes (specifically the 
'server offset SHM1' graph). The only thing that it vaguely matches up 
with is the local GPS long wave due to visibility. But again - spikes 
would certainly not be an expected symptom of visibility!

Are you using a common power supply/source for all three pi's - including 
the ntpviz device?

Pure shots in the dark. I've dealt with intermittent spiking in the long 
ago past, with a resolution that is very unlikely to be related (turning 
off the Monit monitoring application, which isn't part of the default 
install of debian/raspbian)

-- 
Paul Theodoropoulos
www.anastrophe.com