Follow on from: https://bugs.launchpad.net/juju-core/+bug/1611391
presumably because the migration hasn't fully "completed successfully", we are unable to "juju destroy-model A:foo".
However:
1) "juju destroy-model" returns successfully, making you think it has already killed the model, or at the very list it will die soon.
2) no obvious errors in "juju debug-log -m A:controller"
3) no way to kick the migration and abort it, nor complete it, nor forcibly destroy the model. I seem to have to kill the entire controller in order to recover.
I actually have the same problem running
"juju destroy-application elasticsearch"
It doesn't give any indication that it didn't work, but the services still appear to be perfectly happy (they are not in a DYING state).
I would have expected that we might go into a "destroying applications not allowed during a migration", and at least inform the user that we are rejecting their request.
Oh, and I can't run "juju destroy-controller" because it waits for the model to be killed, and I can't run "juju kill-controller" because it does a destroy-controller before it tries to do the rest.
Alexis just ran into this as well found something potentially useful. Although destroy-controller --destroy- all-models was stuck she was able to destroy each model separately and then destroy the controller (without --destroy- all-models) . There might be some interaction with what --destroy- all-models is doing.