Overview
Explains a resiliency mechanism that allows data traffic to continue forwarding using cached information if the control plane connection is lost.
OMP graceful restart
OMP graceful restart is a control plane resiliency mechanism that
-
allows Cisco IOS XE Catalyst SD-WAN devices to continue forwarding data traffic when the Cisco Catalyst SD-WAN Controller is unavailable
-
uses cached OMP information (such as routes, TLOCs, service routes, and policies) to maintain data plane operations, and
-
synchronizes updated network information when the controller connection is restored.
When OMP graceful restart is enabled, both Cisco IOS XE Catalyst SD-WAN devices and Cisco Catalyst SD-WAN Controllers cache OMP information received from their peers. This cache includes OMP routes, TLOC routes, service routes, IPsec SA parameters, and centralized data policies.
How graceful restart events work
Summary
OMP graceful restart enables devices and controllers to maintain forwarding operations and synchronize network state after connectivity is restored.
The key components involved in the process are:
-
Cisco IOS XE Catalyst SD-WAN device: Maintains cached OMP information and forwards data traffic when the controller is unavailable.
-
Cisco Catalyst SD-WAN Controller: Maintains cached OMP information and resumes synchronization when device connectivity is restored.
-
OMP session: Facilitates communication and state synchronization between devices and controllers.
Workflow
These stages describe how OMP graceful restart enables devices and controllers to continue forwarding traffic using cached information during control plane outages and to resynchronize network state when connectivity is restored.
- A device detects loss of OMP connection to a Cisco SD-WAN Controller and continues forwarding data traffic using cached OMP information.
- The device periodically checks whether the controller is available.
- When the controller becomes available again and the OMP session is re-established, the device flushes its local cache and accepts only the new OMP information from the controller.
- The same process occurs in reverse if a Cisco Catalyst SD-WAN Controller loses connection to a device: the controller uses cached information until the device is available again, then updates its state upon reconnection.
Result
Data traffic continues to be forwarded using cached information during control plane outages, and devices/controllers automatically resynchronize when connectivity is restored.
OMP graceful restart timers
Each OMP peer independently configures its graceful restart timer on both Cisco IOS XE Catalyst SD-WAN devices and Cisco SD-WAN Controllers.
For example:
-
If a controller is set to 300 seconds (5 minutes) and a device to 600 seconds (10 minutes), the controller retains OMP routes from the device for 10 minutes (per device's timer), and the device retains routes from the controller for 5 minutes (per controller's timer).
-
The timer value is communicated during OMP session setup and determines how long cached routes are considered valid during peer loss.
When you change OMP graceful restart configuration, the OMP session between Cisco SD-WAN Controller and the device is intentionally reset (flapped). This action withdraws and relearns OMP routes for all address families (such as TLOC, IPv4/IPv6 unicast, IPv4 multicast, etc.) within a few seconds. During this period, Bidirectional Forwarding Detection (BFD) sessions will also flap momentarily. This is the expected behavior.