Removing a unit or application doesn't resolve errors

[Adam Israel]

Scenario:

A deployed application has one or more units in an error state. I call either remove-application or remove-unit to remove it.

What I expect to happen:

The application or unit is removed.

What actually happens:

The unit remains in an error state, with no feedback to the user that the application/unit can't be removed.

The user then has to run `juju resolved --no-retry <unit>` multiple times until it reaches a non-error state, in which case it will be removed.

Removing an application or unit in an error state is not an intuitive process, often leading to confusion ("it doesn't work"), overkill (destroying the model or controller), or non-standard workarounds (scripts to run `juju resolved --no-retry` in a loop).

The `remove-application` and `remove-unit` commands should either resolve units in an error state by default, provide a flag for the user to force this behavior, or prompt the user for how to proceed if a unit in an error state is found.

[Anastasia (anastasia-macmood)]

I thin that this is a duplicate of bug #1814271.

If a unit or application is in an error state, normal 'remove-' unit or application will not work, you'd need to use '--force' available from Juju 2.6.