On rebuild failure, nodes are left unrecoverable

[Chris Jones]

If one calls nova rebuild on a given node and the process fails to complete writing a new image, the node is placed into the ERROR state and it is impossible to recover via the API.

Given the relatively long window between issuing a rebuild command and the process completing, it seems likely that a variety of site-specific faults could cause a transient inability to complete the rebuild. I think it would be better for operators if they had a way to recover from that failure using the API.

To that end, I think that Ironic should not refuse to even attempt a rebuild on a node that is in the ERROR state.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/114281

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/114281
Committed: https://git.openstack.org/cgit/openstack/ironic/commit/?id=8ca29ebb1a78108dcf602891ed42267bbe14df29
Submitter: Jenkins
Branch: master

commit 8ca29ebb1a78108dcf602891ed42267bbe14df29
Author: Chris Krelle <email address hidden>
Date: Thu Aug 14 08:52:31 2014 -0700

    Allow rebuild of node in ERROR and DEPLOYFAIL state

    This patch allows nodes in ERROR or DEPLOYFAIL state to be rebuilt. This
    allows operators have a chance of recovering a node should it go into
    an error state while deploying or rebuilding.

    Change-Id: Ia7d6b3b796357da83fdb4d40f92f18a502956aa2
    Closes-Bug: #1354437