Currently a replication controller is created for each member in Kubernetes. As a result if the pod stops responding, the replication controller will remove the existing pod and create a new one. Here the new pod will get a new pod id. When this happens Stratos will not be able to manage pods because there is no information sent to Stratos about this process. Therefore I think we may need to do avoid using replication controllers and directly create pods until we get assistance from Kubernetes to retrieve this information.