Networking errors on health check

Issue Background

Health check failures have been reported which were viewed as a consequence of particular corporate network settings with webproxies but I have now replicated this issue.

Issue description

On starting b@d there are a lot of

    default: Warning: Connection reset. Retrying...
    default: Warning: Remote connection disconnect. Retrying...

But the VM eventually starts, health checks however fail with the following traceback

Traceback (most recent call last):
  File "/usr/lib/python3/dist-packages/lava_dispatcher/action.py", line 296, in run_actions
    action_max_end_time, args)
  File "/usr/lib/python3/dist-packages/lava_dispatcher/actions/boot/__init__.py", line 422, in run
    ip_addr = dispatcher_ip(self.job.parameters['dispatcher'])
  File "/usr/lib/python3/dist-packages/lava_dispatcher/utils/network.py", line 60, in dispatcher_ip
    iface = gateways['default'][netifaces.AF_INET][1]
KeyError: 2

The errors ceased once I used an ethernet connection for the host (and in vagrant startup) - though I also did an upgrade of vagrant to 2.1.2 originally suspecting something version related.

Once I am back in the codethink mail offices I will check again and see if I can replicate. I ran demos of B@D at ELCE in Oct 2017 with the host using wifi - though in this case the BBB was connected directly to the host via an ethernet cable - not going via a router.

Acceptance criteria

Health checks should work whatever means the host uses to connect to the internet.