System, method and computer program product for recovering from a failure of a computing device. Start up of a first component of the device is monitored and a determination is made whether the first component has started successfully. If so, a second, higher level component of the device is started. Operational data received from the second component is monitored. If the operational data falls outside of an operational boundary, an action is performed on the second component to enable the second component to operate within a preferred operational boundary. If the first component does not start up successfully, a determination is made if start up of the first component is critical to operation of the second component. If so, a corrective action is performed relative to the first component and afterwards, an attempt is made to start up the second component.