[14:38] I tried to get gitlab onto microk8s with juju and it failed miserably. First gitlab charm reqired juju 2.9.0 where stable is 2.8... Then once bootstrapped it remembered there was a controller already from the 2.8 first test, even though mircrok8s got wiped in between... Does juju has a hidden registry somewhere? Then it didn't launch the service on an IP4 address only on an IPV6 but the internal check used IPV4 and [14:38] nothing was there... [14:41] Wasn't the juju idea to make ops code rock solid? In my case it just introduces more problems than anything else. [21:46] juju client stores it's local info about controllers in .local/share/juju/controllers.yaml