bootstrap: cannot interrupt until controller has been started

[Roger Peppe]

If I interrupt (with ^C) a "juju bootstrap" command, it waits for a lot longer than it should and doesn't necessarily interrupt at all.

Example output:

    % juju bootstrap azure/westeurope
    Creating Juju controller "azure-westeurope" on azure/westeurope
    Looking for packaged Juju agent version 2.2-alpha1 for amd64
    ^?Interrupt signalled: waiting for bootstrap to exit
    No packaged binary found, preparing local Juju agent binary
    Launching controller instance(s) on azure/westeurope...
     - machine-0 (arch=amd64 mem=3.5G cores=1)
    Fetching Juju GUI 2.2.7
    Waiting for address
    Attempting to connect to 192.168.16.4:22
    Attempting to connect to 52.174.22.49:22
    Logging to /var/log/cloud-init-output.log on the bootstrap machine
    Running apt-get update
    Running apt-get upgrade
    Installing curl, cpu-checker, bridge-utils, cloud-utils, tmux
    Installing Juju machine agent
    Starting Juju machine agent (service jujud-machine-0)
    Bootstrap agent now started
    Contacting Juju controller at 52.174.22.49 to verify accessibility...
    Bootstrap complete, "azure-westeurope" controller now available.
    Controller machines are in the "controller" model.
    Initial model "default" added.

In this case, I interrupted before the client had even found the binary, but although it acknowledged the interrupt, it never actually aborted and the controller was bootstrapped.

This is a significant issue when accidentally running a bootstrap command with the wrong parameters - it should abort as soon as possible.

[Anastasia (anastasia-macmood)]

Ctrl-C does successfully interrupt and cleanly destroy controller, depending on when in the bootstrap process the signal was issued.
We have a similar bug, https://bugs.launchpad.net/juju/+bug/1618218, which should address interrupt overall.

@Roger,
I'll mark this bug as a duplicate. Thank you for reporting it!