An interesting question for the large-packets draft is whether we want to
support a periodic probing mode, perhaps leveraging the stability draft's
ideas. Or, should it instead only be consistent?
I can see how using BFD to detect an MTU failure and doing the reroute can make
sense, it's forwarding fault detection after all. But IMO probing MTU should
not be done by BFD, unless we're considering rechartering/BFDv2.
Regards,
Reshad (hat on).
On 2018-10-25, 8:58 PM, "Jeffrey Haas" <jh...@pfrc.org> wrote:
On Thu, Oct 25, 2018 at 10:28:52PM +0000, Les Ginsberg (ginsberg) wrote:
> > https://tools.ietf.org/html/draft-ietf-bfd-stability-02
>
> [Les:] I have read this draft - not sure how relevant it is.
Mostly, the idea being that it's possible to do probing in BFD without
necessarily having to commit your entire detection resources to it.
> Naiming had suggested that MTU sized packets need not be sent all the
time but only occasionally -
> and that a failure might not be used to take the BFD session down -
> rather it would be seen as a "soft-failure" and reported separately from
the BFD session state.
> My response was in that context - which it seems was also in your mind in
your BFD Echo proposal.
>
> This, however, seems not to be what Albert has in mind - as he has since
commented that he really wants to have sub-second detection of MTU issues and
> he wants traffic rerouted "immediately".
Right. This is part of the point I'd made earlier in the thread. Some
consumers need immediate failover and consistent MTU probe while others
might be satisfied with periodic probing.
An interesting question for the large-packets draft is whether we want to
support a periodic probing mode, perhaps leveraging the stability draft's
ideas. Or, should it instead only be consistent?
-- Jeff
