Invalid storage prevents machine removal

[Ian Booth]

When adding storage that is invalid (over size limits) juju can no longer remove the machine.

For example:

juju switch amazon
juju bootstrap
juju deploy cs:~chris.macnaughton/trusty/ceph-osd --storage "osd-devices=ebs,2T,1"
juju destroy-service ceph-osd

Juju hangs on destroying the machine that the bad storage was added to.

[Andrew Wilkins (axwalk)]

shouldn't be able to add invalid storage in the first place (lp:1501709), but it certainly should not pin the machine either

[Andrew Wilkins (axwalk)]

This was actually fixed in 2.3-beta1 (or earlier? I *think* that release though.)

I confirmed with:

 - juju bootstrap aws
 - juju deploy postgresql --storage pgdata=ebs,2T
 (wait for machine to be provisioned; storage fails to provision as expected)
 - juju remove-application postgresql
 (postgresql application, unit and machine are all removed; failed storage remains)
 - juju remove-storage pgdata/0
 (pgdata/0 storage is removed from model)
 - juju destroy-controller -y ...

[Paul Gear (paulgear)]

@axw: Will machines in this state be correctly cleaned up after the controller is upgraded to 2.3?

[Andrew Wilkins (axwalk)]

@paulgear I tried to reproduce the issue with 2.0.1, and couldn't. So I don't know. Do you have any affected models? And reproduction steps?