pxc cluster should error the cluster if the nova instances goto error

[Craig Vyvial]

The pxc cluster stays in a BUILDING task even if the nova instances go straight to error until the cluster building poll times out. Usually the cluster build time out is about 5-10 minutes.

We are polling only the service status and we should also look at the instances status.
https://github.com/openstack/trove/blob/master/trove/taskmanager/models.py#L203-L224

This may also apply to the other clustered datastores as well.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/518639

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/518639
Committed: https://git.openstack.org/cgit/openstack/trove/commit/?id=d7286833e1cbd7b1263f97ed38c8606fd6437904
Submitter: Zuul
Branch: master

commit d7286833e1cbd7b1263f97ed38c8606fd6437904
Author: wangyao <email address hidden>
Date: Thu Nov 9 09:37:20 2017 +0800

    Stop polling if nova instances goto error

    According bug description, if the nova instances goto error, the cluster
    building poll still work until timeout. To solve this problem, instance
    task status was added in the cluster building poll detection.
    Also provide a test case to validate the method.

    Change-Id: Id772d5528c93fcd061aa9d2b2e56c89229ad655c
    Closes-Bug: #1516763

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/trove 9.0.0.0b3 development milestone.