Musings around environments, cloud credentials, and the new Juju Controllers

roger peppe roger.peppe at canonical.com
Fri Nov 13 14:01:11 UTC 2015


On 13 November 2015 at 12:58, Rick Harding <rick.harding at canonical.com> wrote:
> The goal is to get the user onto 'best practices' out of the box and
> deploying into the controller environment isn't best practice since it's no
> longer as easily transient. If you bootstrap, deploy something to try it
> out, then realize you want to try something else you can't just destroy that
> single environment. The whole controller has to go down and get restarted.
> The goal is that as soon as you start working it's light weight and easy to
> destroy/restart without a new bootstrap. I think that trade off is worth the
> change for existing users that machine 0 doesn't appear there out of the
> box.

Just as long as we're going into this with open eyes...

Note that this is technically a backwardly incompatible change as it
will break any scripts that bootstrap an environment and deploy a unit
to machine 0.

Also, it may be unexpected that juju bootstrap followed by juju
destroy-environment
leaves a (potentially expensive) machine running.

BTW although you can't easily destroy that single environment, you can
destroy individual services, which should amount to something very
similar.

  cheers,
    rog.



More information about the Juju-dev mailing list