Not all storage is necessarily reachable/zoned for every node, and not every node is capable of hosting every container. It would seem to be necessary to ensure that a set of nodes (all, a specific NodeId, or a tagged group of NodeId(s)) will be able to reach a new volume before issuing the CreateVolumeRequest. CO(s) often know about node-level fault domains but this is somewhat different as it's a storage connectivity level decision. Could perhaps leverage some of the concepts in https://github.com/container-storage-interface/spec/issues/7 to define the groups. Note that this information might need to be a part of CreateVolumeResponse so we could handle the case where the request specified nothing but the CO still needs to know what nodes are actually capable of mounting a volume.