Wait for all success after nova boot with poll

[tobe]

Now we can use nova boot with poll parameter for one instance. But if we want to boot multiple instances, it return whenever the first instance succeeds or fails.

It would be much better for testing to return the result of all instance when boot with max and min parameters.

[Markus Zoeller (markus_z) (mzoeller)]

Confirmed:
The novaclient uses only one instance to show the progress [1] and
doesn't consider the real number of created instances. AFAIK the REST
API to create instances will always return the first instance and not
a full list of all created instances. Blueprints to change that [2]
didn't get implemented. There was a bug about that too (within the
last 12 months) but I don't find it anymore. IIRC we accept this
behavior, but don't pin me down on this.

However, there is the REST API request parameter "return_reservation_id"
which could maybe used in the python-novaclient to list all instances
matching this reservation_id [3]. And then the progress can be polled
per instance. If this would make the testing more reliable it's worth
a shot IMO.

References:
[1] novaclient; Mitaka; create; poll for one instance: https://github.com/openstack/python-novaclient/blob/b80d8cb6e6cd1e86c7dc3c99c3e7d92641c00097/novaclient/v2/shell.py#L591-L592
[2] https://blueprints.launchpad.net/nova/+spec/return-all-servers-during-multiple-create
[3] nova api-ref; V2.1; create multiple instances: http://developer.openstack.org/api-ref-compute-v2.1.html#os-multiple-create-v2.1

[Markus Zoeller (markus_z) (mzoeller)]

@Prathyusha

Thanks for taking over this bug. Please provide a patch for review within the next 2 weeks. In case you won't work on this anymore, set it back to "Confirmed" and remove yourself as assignee.