[j-nsp] BFD/IS-IS wait to re-establish adjacency after failure tweak knob?

[Duane Grant]

i agree that they could make it work, but it's a biggish change.  bfd is an
L3 protocol and below he's asking it to work at L2 and do neighbor
discovery.  the way bfd works now, the registering protocol does or
provides the neighbor discover part and tells bfd.  this infrastructure
isn't available for what Adam wants to do but i agree that it would be
useful and doable.

i would just love to be able to force protocol registration with bfd so i
can relax my protocol timers.





On Fri, May 20, 2016 at 10:10 AM, Saku Ytti <saku at ytti.fi> wrote:

> On 20 May 2016 at 16:51, Duane Grant <duaneogrant at gmail.com> wrote:
> > in every OS I've used, you need a registering protocol to get bfd to
> start,
> > and if you start shutting them down (or remove peer reachability), bfd
> will
> > admindown itself, which causes interesting consequences.
>
> Sure, but this is implementation issue, there is no technical reason
> it couldn't work as Adam described, and in many scenarios, perhaps
> even in the the most common scenario, this is the cheapest and lowest
> convergence way to do it.
>
> Unlike OAM, due to traction BFD has it's implemented in many NPU/ASICs
> directly, which means predictable performance.
>
> --
>   ++ytti
>