kill-controller / destroy-controller prevents reuse of controller name

I'm not able to recreate with tip of master. Can you try bootstrap / kill-controller / bootstrap with what's in master? (I can provide a binary if needed)

Tried again with same results. Maybe it's aws-specific.

$ dpkg-query -W juju2

juju2 2.0-beta2-0ubuntu1~16.04.1~juju1

$ juju kill-controller controller-aws3

$ juju bootstrap controller-aws4 aws

$ juju kill-controller controller-aws4

$ juju bootstrap --debug controller-aws4 aws

2016-03-14 15:46:06 INFO juju.cmd supercommand.go:59 running juju [2.0-beta2 gc go1.6]
2016-03-14 15:46:06 DEBUG juju.cmd.juju.commands bootstrap.go:355 preparing controller with config: map[type:ec2 name:controller-aws4]
2016-03-14 15:46:06 INFO juju.provider.ec2 provider.go:48 opening model "controller-aws4"
2016-03-14 15:46:06 DEBUG juju.environs.configstore disk.go:326 writing cache file
2016-03-14 15:46:06 ERROR cmd supercommand.go:448 cannot create controller "local.controller-aws4" info: model info already exists

Interestingly enough,

$ juju list-models

error: controller local.controller-aws4 not found

$ juju switch controller-gce

$ juju list-models

NAME OWNER LAST CONNECTION
controller-gce admin@local 2016-03-11

$ juju bootstrap controller-aws4 aws

ERROR cannot create controller "local.controller-aws4" info: model info already exists

Can you attach the full output of the sequence of commands you mention above with the --debug flag?

$ juju bootstrap controller-aws4 aws --debug

$ juju kill-controller controller-aws4 --debug

$ juju bootstrap --debug controller-aws4 aws

You can just put it in one pastebin.

http://paste.ubuntu.com/15393483/

And why does my newly-created controller now have the 'local.' prefix added? This must be something new (and, frankly, annoying) in Juju 2.0.

in Juju 2.0 beta2 I mean

I'm running into the same issue now.

This issue appears to be aws specific. The issue is that the kill/destroy of the controller did not compete without an error, so the final step to clean up does not get run.

2016-03-15 17:11:50 DEBUG juju.provider.ec2 ebs.go:439 destroying "vol-ca061c69"
2016-03-15 17:11:50 ERROR cmd supercommand.go:448 destroying storage: destroying volumes: destroying "vol-ca061c69": The volume 'vol-ca061c69' does not exist. (InvalidVolume.NotFound)

So if the environment is not properly able to be killed, it's to be expected that Juju will not want to destroy potentially important information pertaining to the environment. A new bug needs to be opened for the root cause issue. ie I've tested this on other environments and they clean up ok and there's no issue.

New bug https://bugs.launchpad.net/juju-core/+bug/1559844