The Pacemaker Cloud project provides high levels of service availability for high scale cloud deployments. Our approach to high availability is to detect failures, isolate failures, followed by restart of the failed components. When repeated component failures occur the software escalates those failures into failures of higher level components.
Pacemaker-cloud's README describes how to get started running Pacemaker-cloud. Note that the software is currently in development and only suitable for developers at this time.