Abstract: Aspects and implementations of the present disclosure are directed to methods and systems for a network tolerant to a network device that is temporarily absent from the control plane. In one aspect, in general, the system includes an event manager configured to send, to a controller configured to exchange control messages with a network device, a first request to temporarily withdraw the network device from control plane interactions. The event manager then triggers an event at the network device during which the network device is non-responsive to control plane interactions. After determining that the event has completed, the event manager sends the controller a second request to restore the network device. The controller is configured to request, responsive to the first request, a network application to transition to a tolerant state; and to request, responsive to the second request, the network application to transition to a sensitive state.

Abstract: Embodiments are directed towards managing network routes using a centralized service on behalf of a network of a plurality of forwarding devices. The centralized service develops a global view of the network based on information from the forwarding devices. The centralized service computes changes to the routing policy for each of the nodes. The centralized service then generates an operational command schedule that is distributed to the forwarding devices using a “make before break” sequence to ensure that transient black holes do not happen as the new routing policy is being implemented.