Best practices question

[Paul Theodoropoulos]

I recently added a third ntpsec RPI to my little homebrew cluster. I 
have them all peering with each other, along with each having a few 
'outside' stratum one servers in the mix (different ones on each 
device). so, I have

a-ntpsec, which peers with b- and c-ntpsec (along with 'outsiders')
b-ntpsec, which peers with a- and c-ntpsec. As above.
c-ntpsec, which peers with a- and b-ntpsec. As above.

Is this a reasonable setup, or is there a potential for it to amplify 
errors/instability rather than mitigating it? I realize that in time, I 
should find out from my ntpviz graphs what the effect will be, but if 
this is a misguided setup, I'd rather nip it in the bud sooner than later.

All three are in a thermally insulated box, so temperature is beginning 
to stabilize after a day or so since I wired it all up. The graphs 
beyond 12 hours look pretty horrendous at the moment because of this...

https://ntpsec.anastrophe.com/a-1hr/index.html
https://ntpsec.anastrophe.com/b-1hr/index.html
https://ntpsec.anastrophe.com/c-1hr/index.html

All three are RPI3B+, Raspbian 9.9, ntpsec 1.1.4 from repo, gpsd 
3.19-dev from  repo. a- and c- have the same ublox neo-M8T board, b- has 
an adafruit hat. Interestingly, the brand new RPI (c-ntpsec) has a 
significantly higher and less stable frequency offset from the other 
two. I'm wondering if this is a function of the clock crystal being 
brand new rather than a year+ old, or whether it's just the 'luck of the 
draw'.

Photo of the setup https://i.imgur.com/WTGA8xU.jpg - I gotta brag a 
little on my obsessive-compulsive hobby...

-- 
Paul Theodoropoulos
www.anastrophe.com

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ntpsec.org/pipermail/users/attachments/20190625/d0ddea5f/attachment.htm>