Description of problem:

In order to support a resilient architecture we'd like that glance supports following architecture:

1.- Glance instances should be able to  store and retrieve images from multiple ceph clusters (RBD backends).
2.- Glance should be able to store each image on several multiple RBD backends at upload time. These backeds could be defined by setting a list of backend identifiers in default_store.
3.- For each upload destination, glance should add a location metadata for the image using multi-location feature.  
4.- If glance instances fail to fetch an image from the first location, it should try the next one until it's able to get it or tries all available locations.

As the information provided by Red Hat engineering and some testing at lab environment, current status is:

- glance-store is being refactored to enable (1).
- Glance already should failover to alternative location in case of failure as described in (4)


1 and 3 seem to be covered upstream by https://blueprints.launchpad.net/glance/+spec/multi-store

But still lacking:
   - Manage two or more rbd backends effectively.
   - Upload images to more than one glance backend.