When connecting multiple SolidFire Clusters to an openstack instance, the expectation was to see volumes allocation balanced across the clusters. But this balance is not happening and all volumes are being created on the larger cluster. While it is understandable that the larger cluster use would have to reach a point where the clusters had equal capacity before the balancing can occur, using the provisionable space relative to the in use space is not allowing this balancing to occur.
Exposing the capacity in terms of maxUsable space and used (consumed - not allocated) space would provide a better capacity planning based on thin provisioning.
I propose that the driver be updated to take into account three use cases :
1. Capacity planning based on provisioned and provisionable space, therefor not being able to over allocate a cluster.
2. Capacity planning based on consumed space and available space, therefor taking advantage of thin provisioning and allocation based on available consumable space.
3. Capacity planning with consumed space relative to provisionable space. Useful in the case where clusters are the same size.
Then have a configurable that would allow users to choose their option, with a default to the current behavior so as not to break existing consumers of the current use case.
Fix proposed to branch: master
Review: https:/