Cluster instances could report 'ready' when they are not

[Matthew Van Dijk]

The check used by all cluster strategies is wrong. It allows for both RUNNING and BUILD_PENDING. All cluster strategies use BUILD_PENDING to indicate that the instances are cluster-prepare ready. RUNNING is bad as it can be reported before the instance's prepare even starts, causing the cluster-prepare to start too early and fail.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/206115

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/206115
Committed: https://git.openstack.org/cgit/openstack/trove/commit/?id=76c81c4625648608d7873ede630e3f727b8d2fd6
Submitter: Jenkins
Branch: master

commit 76c81c4625648608d7873ede630e3f727b8d2fd6
Author: Matt Van Dijk <email address hidden>
Date: Mon Jul 27 11:29:55 2015 -0400

    Cluster instances could falsely report 'ready'

    The check used by all cluster strategies is wrong. It allows for both
    RUNNING and BUILD_PENDING. All cluster strategies use BUILD_PENDING to
    indicate that the instances are cluster-prepare ready. RUNNING is bad as
    it can be reported before the instance's prepare even starts, causing
    the cluster-prepare to start too early and fail.

    The fix is to remove RUNNING from the list of cluster-ready states. It
    has been verified that all datastores set instances to BUILD_PENDING
    when used for clusters.

    Change-Id: I048e2b76fda59a25928c6c6d3751235df12d7369
    Closes-bug: #1478603