Sorry to hear that you’re encountering trouble, Jeremy.
The general answer is to use
juju run --application ntp -- <command>, where
<command> is the shell command that you want to execute. That runs
<command> on all 13 units in parallel.
The problem is, I don’t know what
<command> should be. My understanding is that there are two utilities with the same name:
ntpmon is part of NTPsec, which doesn’t have any alerting functionality from what I can see.
ntpmon is bundled within the
ntp charm. Its alertings thresholds seem to be percentiles, rather than the hard coded “10” that you mention
Perhaps @paulgear might have some suggestions?
How have you done this?
It sounds like a good idea would be to add stratum 2 time servers. That should prevent drifts like this from occurring.