All of the prerequisites for supporting multiple VM providers should be
in place. This has been tested with rackspace legacy, rackspace nova,
and hpcloud.
The scripts now use novaclient instead of libcloud. The old v1_0 code that
was removed from novaclient is added here for as long as we continue to
use rackspace legacy. It's slightly modified to handle some operational
considerations (such as cache-busting), and to integrate with the current
version of novaclient. We can remove it when it's no longer needed.
Machines are now generated from snapshot images created from per-provider
base images, this lets us specify, eg, oneiric and precise images from each
provider. Setup scripts take the provider name as an argument (so each
provider in Jenkins can have its own job for easier monitoring). The fetch
script takes the base image name (eg, "oneiric") as an argument and gets
the oldest matching node from any provider.
Snapshot images are created from scratch each time; no more long-running
template hosts.
Devstack fixed network set to something that doesn't collide with hpcloud.
Min_ram is now configurable per-base-image (so we can request servers with
a certain amount of ram for each image (in case an image has no swap, or
otherwise needs more ram)).
SKIP_DEVSTACK_GATE_PROJECT added to the gate script to make testing the
script itself during development easier.
More robust detection of image URLs in the image update script.
On a running devstack node, before running devstack, check to see if there
is swap space. If not, assume we're on HPCloud and unmount /mnt and use
it for swap.
Change-Id: I782e1180424ce0f3c7b69a3042eccc85b2b50389
3435b5a060