If a link goes down and affects an EVC, the consistency check will keep trying to activate/redeploy that EVC on every execution: sending flow_mod to remove the EVC on the UNI devices; submitting a request to pathfinder. Having a few hundred EVCs and a dozen switches can result in unnecessary waste of the controller resources by running such an unnecessary trial.
We could somehow optimize this process while still having this check to ensure the EVC will eventually converge.
We should keep in mind that mef_eline already listens to link_up events and try to activate circuits that would benefit from that link.
@italovalcy, if that's OK, let's shift this to 2023.2 just so the strategies can be better discussed. Considering all the current priorities this wouldn't have enough time to be properly discussed and e2e tested.
Hi,
If a link goes down and affects an EVC, the consistency check will keep trying to activate/redeploy that EVC on every execution: sending flow_mod to remove the EVC on the UNI devices; submitting a request to pathfinder. Having a few hundred EVCs and a dozen switches can result in unnecessary waste of the controller resources by running such an unnecessary trial.
We could somehow optimize this process while still having this check to ensure the EVC will eventually converge.
We should keep in mind that mef_eline already listens to
link_up
events and try to activate circuits that would benefit from that link.