libvirt Spice console requests should be able to fall back to novnc

[Rafi Khardalian]

Using master or Grizzly, there is currently no method for a Spice console request to automatically fallback to novnc. This is important because otherwise there is no migration path if deployments decide to switch from novnc to Spice as their default.

The reason is that the enabling of Spice cannot occur until the instance in question is destroyed, XML redefined enabling Spice, then restarted. Thus, upon deployment, existing instances will only have VNC consoles enabled, where new instances will have VNC + Spice. If the deployment preference is Spice, users and/or Horizon should simply be able to ask Nova for a Spice console and get back a novnc URL if Spice isn't ultimately available.

We can make this fallback from VNC to Spice a configuration option, since the behavior may not be desired in some cases.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/33737

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/33737
Committed: http://github.com/openstack/nova/commit/cc8021df5fd04c0872bd0dece552c0e7e84b6a16
Submitter: Jenkins
Branch: master

commit cc8021df5fd04c0872bd0dece552c0e7e84b6a16
Author: Rafi Khardalian <email address hidden>
Date: Thu Jun 20 00:04:07 2013 +0000

    Raise exceptions when Spice/VNC are unavailable

    Bug 1192724

    Both Spice and/or VNC may be enabled from the perspective of compute
    configuration but their actual availability can vary based on
    whether either was enabled/disabled at the time the instance was
    launched. This patch raises a proper exception in this case, such
    that clients (including Horizon) can either message the user or fall
    back.

    Change-Id: If06e726e606f4a27c8ec4617400b728ce7112e3b