Juju CI seesa lot of evidence that destroy-environment does not do everything it needs to to complete. CI sees many cases where juju claims the env is already bootstrapped., sometime after destroy-environment --force was called. The test job fails, but the next might pass. I suspect that Azure is still cleaning up resources
We commonly find azure jenv files left behind. but even when there isn't, we see the "already bootstrapped" message.
The azure dashboard shows that no machines or services are running, but the networks do exist. Storage still has a container with state files in it :(
Manually deleting these can help, but we still need to wait about 10 minutes before we run the job again, otherwise we see the "already bootstrapped" message. We cannot always delete. The dashboard often fails to delete networks and blobs.
CI may change azure testing to create a different env name to ensure resources from the last test are not reused, but we then run the risk of using up the limited resources azure provides.
The 1.18.4 release was stalled by this bug.
$ juju --show-log destroy-environment --force curtis-azure
2014-06-03 14:48:08 INFO juju.cmd supercommand.go:302 running juju-1.
WARNING! this command will destroy the "curtis-azure" environment (type: azure)
This includes all machines, services, data and other resources.
Continue [y/N]? y
2014-06-03 14:48:23 ERROR juju.cmd supercommand.go:305 cannot delete the environment's virtual network: asynchronous operation failed: BadRequest - An error occurred when setting the network configuration: The virtual network juju-curtis-
My general reaction is "just fucking die". The network cannot be deleted from the azure dashboard either.