Re: [RTG-DIR] Rtgdir early review of draft-ietf-rtgwg-vrrp-p2mp-bfd-08

[Greg Mirsky <[e-mail]>]

Hi Emmanuel,
thank you for your thorough review and thoughtful suggestions. Please find
my notes below tagged by GIM>>. Attached, please find the new working
version of the draft.

Regards,
Greg

On Wed, Mar 13, 2024 at 12:46 AM Emmanuel Baccelli via Datatracker <
[e-mail]> wrote:

> Reviewer: Emmanuel Baccelli
> Review result: Has Nits
>
> Hello,
>
> I've been selected as Routing Directorate (early) reviewer for this draft.
>
> I have a few nits (nothing major) and a couple of suggestions.
>
> Some of my comments might come through as pedantic -- mostly due to my
> superficial prior knowledge concerning VRRP!
>
> # Abstract:
>
> suggested change/clarification
> "...sub-second convergence of the Active router and..."
> => "...sub-second convergence for the process determining the Active router
> and..." or something equivalent.
>
GIM>> Thank you for your suggestion. Accepted.

>
> # Section 1:
>
> suggested change/clarification
> "this document demonstrates how... can enable faster detection..."
> => "this document specifies fast transition to a new Active router, upon
> detection of..." or something equivalent.
>
GIM>> VRRPv3 allows for sub-second switchover to a new Active router
although using high-rate protocol's Hello. This proposal describes an
alternative mechanism, i.e., p2mp BFD, to support that functionality. It
seems like the innovation the draft provides is in the way to support the
essential functionality using a light-weight BFD mechansim. Perhpas the
current wording is acceptable.

>
> # Section 2:
>
> "Supporting sub-second mode... in the data plane may prove challenging"
> => Would be best to hint at the main reason why (costs in terms of control
> traffic overhead?).
>
GIM>> Clarified with the following update:
NEW TEXT:
   Supporting sub-second mode
   for VRRPv3 [I-D.ietf-rtgwg-vrrp-rfc5798bis] for all these roles
   without specialized support in the data plane may prove challenging
   because of the increased load on the control plane.

>
> "BFD already has many implementationq based on HW"
> => Cite at least one implementation, if possible?
>
GIM>> By now BFD, to the best of my understanding, supported by all HW
platforms. Naming one and not mention others seems like undesirable and
devisive. WDYT?

>
> # Section 3:
>
> My Discriminator => cite RFC5880 upon first use of this term in the doc ;)
>
GIM>> Thanks! Added.

>
> "... starts transmitting BFD control packets with VRID as a source IP
> address
> and ..." => it is unclear how VRID (1 Byte) can be used as IP address. Can
> you
> rephrase/clarify?
>
GIM>> You're correct. Based on the comments by Joe Clarke
<https://mailarchive.ietf.org/arch/msg/rtgwg/Wd7agbSfnSIxfjs_jSkNWt6ZjMQ/>,
the new -08 version is as follows:
   The Active router, configured to use p2mp BFD to support faster
   convergence of VRRP, starts transmitting BFD control packets with
   IPvX address associated with the Virtual Router
   [I-D.ietf-rtgwg-vrrp-rfc5798bis] as a source IP address and the
   locally allocated value as the value of the My Discriminator field
   ([RFC5880]).

WDYT?

>
> "... when a backup router detects failure of the Active router, ..."
> => using which mechanism/RFC ? I suggest citing it explicitly
>
GIM>> Added reference to Section 5.11 RFC 8562 that defines how detection
time is calculated for the MultipointTail in p2mp BFD session.

>
> "... it reevaluates its role as VRID."
> => it is unclear how this happens exactly. If this is intentionally left
> unspecified as implementation-dependent, I suggest to say it explicitly in
> the
> doc.
>
GIM>> That passage was also changed as a result of addressing comments from
Joe Clarke. I hope that the new text is clearer:

... it re-evaluates its role in the Virtual Router.


> "... the new Active router MUST select My Discriminator and..."
> => it is unclear which discriminator is meant here. Do you mean the value
> locally allocated (as it was still Backup router)?
>
GIM>> Thank you for pointing it out to me. It is unclear. I propose the
following update:
NEW TEXT:
   As a result, the Backup router may become the Active router
   of the given Virtual Router or continue as a Backup router.  If the
   former is the case, then the new Active router MUST select its new My
   Discriminator value, include that value in the VRRP packet to
   bootstrap a new p2mp BFD session, and start transmitting p2mp BFD
   control packets using the Active router IP address as the source IP
   address for p2mp BFD control packets and its new My Discriminator
   value.

Is it clearer now?


> # Section 5:
>
> "... to accelerate detecting a failure that affects VRRP"
> => it is unclear what in the doc accelerates *detecting* a failure. I
> suggest a
> rephrase such as "...to accelerate transition to a new Active router upon
> detection of BFD failure" or something equivalent.
>
GIM>> I think that the switchover, after the detection of the network
failure, is implementation-dependent.  The document is aimed to describe
   "an alternative way, to the one defined in
   [I-D.ietf-rtgwg-vrrp-rfc5798bis], to accelerate detecting a failure
   that affects VRRP functionality using p2mp BFD.  The operation of
   either protocol is not changed."