I've seen this 3 times now, even filed a bug earlier today - LP:1797823. The most recent happening to a bunch of environments over the weekend, when no one from IS is around to run things - see pastebin of nagios output for a bunch of environments - https://pastebin.canonical.com/p/wQqr2y9Mbb/
We have a cron job that periodically dumps the 'juju status' output for monitoring. From what I've seen, the juju client somehow overwrite ~/.local/share/juju/models.yaml with something very bare:
| controllers:
| prodstack-is: {}
We've only started seeing this since upgrading client+controller from 2.3.9 to 2.4.3, if this helps.
'juju switch <model>' fixes this and writes out a proper models.yaml with 'current-model' but seems this is temporary until something happens, perhaps to do with issues with the controller? Still, the client shouldn't be overwriting models.yaml overwriting 'current-model' unless someone runs 'juju switch'.
I've seen this 3 times now, even filed a bug earlier today - LP:1797823. The most recent happening to a bunch of environments over the weekend, when no one from IS is around to run things - see pastebin of nagios output for a bunch of environments - https:/ /pastebin. canonical. com/p/wQqr2y9Mb b/
We have a cron job that periodically dumps the 'juju status' output for monitoring. From what I've seen, the juju client somehow overwrite ~/.local/ share/juju/ models. yaml with something very bare:
| controllers:
| prodstack-is: {}
We've only started seeing this since upgrading client+controller from 2.3.9 to 2.4.3, if this helps.
'juju switch <model>' fixes this and writes out a proper models.yaml with 'current-model' but seems this is temporary until something happens, perhaps to do with issues with the controller? Still, the client shouldn't be overwriting models.yaml overwriting 'current-model' unless someone runs 'juju switch'.