Protocol convergence timer LACP and BFD

Convergence is always challenging task, tunning it to best value is always difficult part. Tunning it is on going process.

According to me the goal of the convergence is to detect the network changes and programing data plane quickly and efficiently. Many times to achieve that we end up turning too many knobs which turns out to be expensive.

In this blog post we will review tuning of LACP or BFD timers or rather the discussion is which one should be aggressive over another one?

I would prefer to keep LACP more aggressive (to fast periodic) and BFD timers littler longer then the LACP. Lets discuss that in following scenario with aggregate ethernet(AE) link.

Let’s say if we have member links A, B and C in AE bundle and due to vendor specific algorithm link A is chosen for protocols traffic including BFD protocol.

We have BFD protocol timers are more aggressive then the LACP timer and now if link A were to report some issue, there is a chance that the BFD will report the time out before the LACP even detect the link level issue. BFD going down would trigger its client protocol to report adjacency down and it will trigger the network convergence even though the link B and C are up and running with out any errors.

In above scenario, if LACP timers were much aggressive then the BFD, AE bundle would detect the issue with link A and it would take it out from bundle and move the protocol generated traffic to link B or C before BFD detect the timeout.

This is just one such observation which could help in avoiding unwanted convergence in the network.