Re: draft-ietf-bfd-vxlan IESG status

Benjamin Kaduk <kaduk@mit.edu> Wed, 17 June 2020 04:42 UTC

Return-Path: <kaduk@mit.edu>
X-Original-To: rtg-bfd@ietfa.amsl.com
Delivered-To: rtg-bfd@ietfa.amsl.com
Received: from localhost (localhost [127.0.0.1]) by ietfa.amsl.com (Postfix) with ESMTP id 0440C3A0E81; Tue, 16 Jun 2020 21:42:51 -0700 (PDT)
X-Virus-Scanned: amavisd-new at amsl.com
X-Spam-Flag: NO
X-Spam-Score: -1.897
X-Spam-Level:
X-Spam-Status: No, score=-1.897 tagged_above=-999 required=5 tests=[BAYES_00=-1.9, RCVD_IN_MSPIKE_H4=0.001, RCVD_IN_MSPIKE_WL=0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, URIBL_BLOCKED=0.001] autolearn=ham autolearn_force=no
Received: from mail.ietf.org ([4.31.198.44]) by localhost (ietfa.amsl.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id WooeECLog8Jm; Tue, 16 Jun 2020 21:42:48 -0700 (PDT)
Received: from outgoing.mit.edu (outgoing-auth-1.mit.edu [18.9.28.11]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ietfa.amsl.com (Postfix) with ESMTPS id E19413A0E7F; Tue, 16 Jun 2020 21:42:47 -0700 (PDT)
Received: from kduck.mit.edu ([24.16.140.251]) (authenticated bits=56) (User authenticated as kaduk@ATHENA.MIT.EDU) by outgoing.mit.edu (8.14.7/8.12.4) with ESMTP id 05H4gexv021916 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-GCM-SHA384 bits=256 verify=NOT); Wed, 17 Jun 2020 00:42:43 -0400
Date: Tue, 16 Jun 2020 21:42:40 -0700
From: Benjamin Kaduk <kaduk@mit.edu>
To: Jeffrey Haas <jhaas@pfrc.org>
Cc: Greg Mirsky <gregimirsky@gmail.com>, iesg@ietf.org, draft-ietf-bfd-vxlan@ietf.org, rtg-bfd WG <rtg-bfd@ietf.org>, bfd-chairs@ietf.org
Subject: Re: draft-ietf-bfd-vxlan IESG status
Message-ID: <20200617044240.GF11992@kduck.mit.edu>
References: <20200127221705.GB17622@pfrc.org> <20200616211057.GA21373@pfrc.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20200616211057.GA21373@pfrc.org>
User-Agent: Mutt/1.12.1 (2019-06-15)
Archived-At: <https://mailarchive.ietf.org/arch/msg/rtg-bfd/iMjAshINtbxtDrX542C6ADIXg7s>
X-BeenThere: rtg-bfd@ietf.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: "RTG Area: Bidirectional Forwarding Detection DT" <rtg-bfd.ietf.org>
List-Unsubscribe: <https://www.ietf.org/mailman/options/rtg-bfd>, <mailto:rtg-bfd-request@ietf.org?subject=unsubscribe>
List-Archive: <https://mailarchive.ietf.org/arch/browse/rtg-bfd/>
List-Post: <mailto:rtg-bfd@ietf.org>
List-Help: <mailto:rtg-bfd-request@ietf.org?subject=help>
List-Subscribe: <https://www.ietf.org/mailman/listinfo/rtg-bfd>, <mailto:rtg-bfd-request@ietf.org?subject=subscribe>
X-List-Received-Date: Wed, 17 Jun 2020 04:42:51 -0000

Hi Jeff,

A big "thank you!" for continuing to track the many issues, I think I have
a sense for the effort involved, but it makes a huge difference.
Since I was mentioned, a couple comments inline.

On Tue, Jun 16, 2020 at 05:10:57PM -0400, Jeffrey Haas wrote:
> [Apologies on further delay.  The best way to cause unexpected work is to
> offer a personal deadline to have something done.]
> 
> Greg and the IESG,
> 
> This update is vs. version -12 of the draft.
> 
> General summary: Almost ready to go.  Multiple issues are resolved.  Pending
> items flagged here should be addressed.  Once addressed, we'll ask the IESG
> to take this up again.
> 
> On Mon, Jan 27, 2020 at 05:17:05PM -0500, Jeffrey Haas wrote:
> > Much like the BFD Working Group discussion on the BFD for vxlan feature, the
> > IESG review for the draft has reached a stage where it is difficult to
> > determine what the related actions are.  (IESG take note for tools
> > discussion!)
> > 
> > This email is an attempt to kick the conversation back into gear.
> > 
> > My notes here are based on the current status of the document tracked here:
> > https://datatracker.ietf.org/doc/draft-ietf-bfd-vxlan/ballot/
> > 
> > My comments on the draft are based on the -10 version of the draft as
> > currently published.
> > 
> > ---
> > 
> > Open Issue 1: Discussion on TTL/Hop Limit = 1
> > 
> > Proposed Action: Greg has proposed text he will send to the working group
> > suggesting GTSM procedures be utilized.  The expected concern is how this
> > impacts existing implementations.
> 
> This issue is resolved.
> 
> > 
> > ---
> > 
> > Open Issue 2: Document Status should be Informational rather than Proposed
> > Standard.
> > 
> > Proposed Action: Greg should make the document Informational.  Prior WG
> > discussion suggested that we don't really care what level it should be at,
> > and had actually requested IESG guidance long ago via our AD.
> 
> This action is pending.
> 
> > ---
> > 
> > Open Issue 3: dst IP/MAC assignment procedures for inner VXLAN headers.
> > (DISCUSS via Benjamin K.)  Specifically per-VNI form rather than strictly
> > VTEP-to-VTEP mode.
> > 
> > Issue Comment 1 (Benjamin K.): This is "a namespace grab in what is
> > essentially the tenant's namespace".
> > 
> > Issue Comment 2 (Jeff H.): Joel Halpern flagged this repeatedly as well as
> > part of directorate review.
> > 
> > Issue Comment 3 (Benjamin K.): "management VNI does not suffer from this
> > namespacing issue".
> > 
> > Issue Comment 4 (Jeff H./Benjamin K.): The concept of a "management VNI" is
> > not supported by existing standards work, but is accepted as a common
> > implementation behavior.
> > 
> > Issue Comment 5: A significant exploration of this set of issues is
> > documented in the following thread:
> > https://mailarchive.ietf.org/arch/msg/rtg-bfd/SfXfu3pCh9BxaRFrXbEOgGt6xjE/
> > 
> > Proposed Action: Limit the Internet-Draft's applicability to verifying
> > connectivity to the management VNI.  "All other uses of the specification to
> > test toward other vxlan endpoints are out of scope."
> > 
> > In reviewing the thread, my reading of the comments from Santosh, Anoop, and
> > Dinesh are effectively "don't break existing implementations".  There is
> > acknowledge among those in the discusssion that numbering space collisions
> > between the protocol codepoints chosen to run as endpoints for the BFD for
> > vxlan session and the tenant space are undesirable.  It is generally agreed
> > in the thread (IMO) that for the "management VNI" case that this is not
> > problematic, although the details of provisioning are still specific to the
> > implementation.  
> > 
> > By setting the case aside where a test to a specific VTEP may have tenant
> > namespace collisions, the document can be cleaned of a lot of unnecessary
> > edge cases that are difficult to generally resolve.  Implementations that
> > may choose to permit sessions to non-management VNIs will have need to
> > resolve how to deal with collisions.  
> 
> The suggested action was taken.  The document now refers only to the
> management VNI, and offers a default value for that VNI number.
> This should resolve a significant number of issues.

Indeed, I think that simplifies things on many fronts.

> > 
> > ---
> > 
> > Open Issue 4: "multicast service node" text (COMMENT via Benjamin K.)
> > 
> > Proposed Action: Incorporate suggested text from Benjamin K. to clarify text
> > in -10.
> 
> This issue is resolved.
> 
> > ---
> > 
> > Open Issue 5: Comma parsing issue (COMMENT via Benjamin K.)
> > 
> > Proposed Action: Accept Benjamin's suggested changes.  (RFC Editor will win
> > the day here though!)
> 
> This action is pending.
> 
> > ---
> > 
> > Open Issue 6: "Section 3, MUST NOT be forwarded to a VM" (COMMENT via
> > Benjamin K.)
> > 
> > Proposed Action: The fate of this issue is tied to Open Issue 3.
> > If the proposal is limited solely to management VNI, this text is not
> > relevant and may be deleted.
> 
> This action is still open.  
> 
> > ---
> > 
> > Open Issue 7: "::FFFF:7F00:0/104 IPv6 range" (COMMENT via Benjamin K.)
> > 
> > Proposed Action: I believe this issue's fate is similarly tied to Open Issue 3.
> > If the proposal is limited solely to management VNI, this text is not
> > relevant and may be deleted.
> 
> This action is still pending. 
> 
> > ---
> > 
> > Open Issue 8: "Section 4: MUST ensure that the BFD Control packet is not
> > forwarded to a tenant but is processed locally at the remote VTEP" (COMMENT
> > via Benjamin K.)
> > 
> > Proposed Action: I believe this issue's fate is similarly tied to Open Issue 3.
> > If the proposal is limited solely to management VNI, this text is not
> > relevant and may be deleted.
> 
> This action is still pending and perhaps requires further discussion.
> 
> Benjamin's point was:
> "This has to be 100% reliable, and I think we need to provide some
> example mechanism that has that property even if we don't mandate that
> it be the only allowed mechanism."
> 
> The motivation here was that when this specification was intended to address
> a fully generalized BFD for vxlan for arbitrary VNIs, there was a strong
> need to say "do not forward the BFD traffic to the tenant".
> 
> For the now default scenario of only the management VNI, this text may be
> adequate.  Benjamin should look at the current document and decide if the
> scenario is still of concern.

I will defer looking at the document until the other couple things are
ready, but I'm pretty confident that this should be fine.  The normal
mechanisms used to keep the management VNI's traffic out of tenant systems
will suffice to keep BFD traffic out of tenant systems, so there shouldn't
be a need to specify additional mechanism here.

> > ---
> > 
> > Open Issue 9: "Destination MAC: This MUST NOT be of one of tenant's MAC
> > addresses." (COMMENT via Benjamin K.)
> > 
> > Proposed Action: I believe this issue's fate is similarly tied to Open Issue 3.
> > If the proposal is limited solely to management VNI, this text is not
> > relevant and may be deleted.
> 
> In -12, we now have the following text:
> "Destination MAC: since a Management VNI is the VNI that does
> not have any tenants, the value of this field is not analyzed
> by the receiving VTEP."
> 
> For the management VNI, this text is true.  However, it likely requires a
> value that is described in the specification.
> 
> Proposed solution: A MAC value should be chosen that is well known and the
> text would become:
> 
> "Destination MAC: A Management VNI, which does not have any tenants, will
> have no dedicated MAC address for decapsulated traffic.  The value X:X:X:X:X

s/no/one/ or s/no/a/?

> SHOULD be used in this field."
> 
> SHOULD might need to be MUST.  Since a partial motivation for permitting the
> flexibility in the specification to NOT use the management VNI is desired,
> MUST might be inappropriate.

IIRC there was mentioned way upstream a range of MAC addresses that we can
make allocations from and the procedure to do so.  Personally I would
prefer MUST but I have a creative enough imagination that I won't object to
SHOULD.

> > ---
> > 
> > Open Issue 10: "The details of how the MAC address is obtained are outside
> > the scope of this document." (COMMENT via Benjamin K.)
> > 
> > Proposed Action: None.  Reason 1: If we go with only the management VNI,
> > provisioning remains an easy answer.  Reason 2: If we go with VNI-to-VNI
> > mode, it's not unreasonable for the environment to claim a MAC address.
> > This is no different than a switch itself.  Collisions would be handled via
> > updated configuration.
> 
> This issue is resolved by only describing the management VNI.
> 
> > ---
> > 
> > Open Issue 11: "dst IP header MUST NOT be one of tenant's" (COMMENT via Benjamin
> > K.)
> > 
> > Comment 1 (Jeff H.): The loopback range as a destination would serve to
> > catch BFD traffic in either VNI-to-VNI or VTEP-to-VTEP mode.  I think this
> > is more clearly understood after the IESG reviewed the existing mechanisms
> > using the loopback address range in existing RFCs.
> > 
> > Proposed Action: I believe this issue's fate is similarly tied to Open Issue 3.
> > If the proposal is limited solely to management VNI, this text is not
> > relevant and may be deleted.
> 
> In the current text for -12, the host loopback range serves to provide the
> necessary mechanism.  Given the restricted scope of the the management VNI,
> I believe the unchanged text is sufficient.

That sounds right; I'm pretty sure that concerns over destination IP only
came into play when we were using non-management VNIs.

Thanks again,

Ben

> > ---
> > 
> > Open Issue 12: "Section 5, BFD dst mac collision with tenant" (COMMENT via
> > Benjamin K.)
> > 
> > Proposed Action: I believe this issue's fate is similarly tied to Open Issue 3.
> > If the proposal is limited solely to management VNI, this text is not
> > relevant and may be deleted.
> 
> I believe the text in -12 describing only the management VNI resolves this
> issue.
> 
> > ---
> > 
> > Open Issue 13: "The UDP destination port and the TTL of the inner IP packet
> > MUST be validated to determine if the received packet can be processed by
> > BFD." (COMMENT via Benjamin K.)
> > 
> > Proposed Action: Provide reference to RFC 5880/5881 sections covering
> > existing BFD procedure.  Do not copy and paste from them.
> 
> The text in -12 is:
> 
> "Validation of TTL / Hop Limit of the inner IP packet, as long as the
> related considerations for BFD control packet demultiplexing and
> authentication, is performed as described in Section 5 [RFC5881]."
> 
> This text is slightly awkward English, but I believe covers the intent.
> I suggest rewording this to:
> 
> "The received packet's inner IP payload is then validated according to
> Sections 4 and 5 in [RFC5881]."
> 
> > ---
> > 
> > Open Issue 14: "nits ... then the BFD session" (COMMENT via Benjamin K.)
> > 
> > Proposed Action: Accept grammar correction.
> 
> This issue is no longer relevant in -12.
> 
> > ---
> > 
> > Open Issue 15: "Section 6" (COMMENT via Benjamin K.)
> > 
> > Proposed Action: I believe this issue's fate is similarly tied to Open Issue 3.
> > If the proposal is limited solely to management VNI, this text is not
> > relevant and may be deleted.  In particular, this section attempts to
> > justify VNI-to-VNI mode poorly.  
> 
> This issue is no longer relevant in -12.
> 
> > ---
> > 
> > Open Issue 16: "Section 9" regarding mis-forwarding/filtering of BFD traffic
> > toward tenant (COMMENT via Benjamin K.)
> > 
> > Proposed Action: I believe this issue's fate is similarly tied to Open Issue 3.
> > If the proposal is limited solely to management VNI, this text is not
> > relevant and may be deleted.
> 
> I believe the new text, describing only the management VNI, obviates this
> comment.
> 
> > ---
> > 
> > (Alvaro's DISCUSSes are covered by the above.)
> > 
> > (Alvaro acknowledged that his COMMENTS were cleared on December 25, 2019)
> > 
> > (Eric V's DISCUSS points are covered by prior open points regarding:
> >  - TTL (see Open Issue 1)
> >  - Mapped IPv6 addresses were covered in discussion with IESG about existing
> >    RFC behavior for this range.
> >  - The mismatch between document IANA action and shepherd writeup is an
> >    artifact of document changes since the shepherd writeup had happened.
> >    The document currently has no open IANA actions.
> >  - Section 9 issues about TTL were addressed in -09 of the document.)
> > 
> > ---
> > 
> > Open Issue 17: "RFC 5881 (BFD) states that it applies to IPv4/IPv6 tunnels,
> > may I infer that this document is only required to address the Ethernet
> > encapsulation ? I.e.  specifying the Ethernet MAC addresses?" (COMMENT via
> > Eric V.)
> > 
> > Comment 1 (Jeff H.): RFC 5881 addresses single-hop "that is associated with
> > an incoming interface".  vxlan requires additional demultiplexing based on
> > packet contents and thus the comment is not fully applicable.  This document
> > (draft-ietf-bfd-vxlan) is intended to cover the vxlan protocol encapsulation
> > for BFD.
> > 
> > Proposed Action: No action required.
> 
> No action was taken.
> 
> > ---
> > 
> > Open Issue 18: "BFD session per VXLAN VNI" (COMMENT via Eric V.)
> > 
> > Proposed Action: I believe this issue's fate is similarly tied to Open Issue 3.
> > If the proposal is limited solely to management VNI, this text is not
> > relevant and may be deleted.
> 
> This issue should be resolved by only describing the management VNI in -12.
> 
> > ---
> > 
> > Open Issue 19: "Section 4...FCS" (COMMENT via Eric V.)
> > 
> > Proposed Action: Accept suggested change to "Outer Ethernet FCS"?
> 
> This action is pending.
> 
> > ---
> > 
> > Open Issue 20: "using the src mac as the dst mac" (COMMENT via Eric V.)
> > 
> > Proposed Action: ?  I'm unclear what the proposal and comment is here.
> 
> It is unclear what action was requested.
> 
> > ---
> > 
> > (TTL issues noted by Eric V. addressed in Open Issue 1.)
> > 
> > ---
> > 
> > Open Issue 21: "throttling of BFD control packets" (COMMENT via Roman D.)
> > 
> > Proposed Action: The section on throttling is written in a confusing manner
> > and is in need of a re-write.
> > 
> > In particular, what's unclear is what is doing the throttling and why?  If
> > the comment is intended to say that some forms of rate-limiting of the vxlan
> > traffic between two systems is in place that it may impact BFD, it should
> > say that.  And perhaps once said, omit giving "advice".  "If it hurts, don't
> > do that."
> 
> These issues were addressed by referring to existing security considerations
> in RFCs 5880, 5881, and 7348.
> 
> > ---
> > 
> > (COMMENTS from Roman D. addressed in -10 and earlier:
> >  - citing specific security considerations applicability
> >  - nits
> > )
> > 
> > (COMMENT from Suresh K. covered in open issues above.)
> > 
> > (COMMENT from Warren K. regardig loopback network range discussed above.)
> > 
> > ---
> > 
> > Open Issue 22: "terminology isn't" (COMMENT via Warren K.)
> > 
> > Proposed Action: Either rename the section "acronyms used in this document"
> > or expand the section to cover the terminology.
> 
> This action is still pending.
> 
> > ---
> > 
> > (Mirja K. indirects a number of issues to "See Olivier's TSV-ART review",
> > which is present in this message:
> > https://mailarchive.ietf.org/arch/msg/rtg-bfd/y3xDkvpT-ZodhcaBRHNOSDVByA8/)
> > 
> > Open Issue 23: "follow same lookup path needs more explanation"
> > 
> > Proposed Action: Add a sentence explaining that this is to ensure that the
> > encapsulated BFD traffic requires following the equivalent data path to
> > protect the resource"
> 
> Section 5 now contains:
> "BFD packets MUST be encapsulated and sent to a remote VTEP as
> explained in this section.  Implementations SHOULD ensure that the
> BFD packets follow the same forwarding path as VXLAN data packets
> within the sender system."
> 
> I believe this addresses the issue.
> 
> > ---
> > 
> > Open Issue 24: "Discuss ECMP considerations" (TSV-ART via Olivier B.)
> > 
> > Comment 1 (Jeff H.): I believe this came up in the message thread, but base
> > BFD also has similar unclear ECMP behaviors.  The working group has avoided
> > trying to standardize anything regarding ECMP since it gets very
> > implementation specific.  Some vendors will go out of their way to do things
> > to mitigate ECMP considerations when BFD is in place; others simply ignore
> > it.
> > 
> > Proposed Action: Unclear.  None?
> 
> No action has been taken here.  I believe this is appropriate.
> 
> > ---
> > 
> > (Minor issues report by Olivier B. that have been addressed in -10 or
> >  earlier:
> >  - p2p vxlan tunnel wording
> >  - VNI has been added to section 2.1
> >  - "figure 1 could take less space" - not addressed.
> >  - section 4 flattened to remove unnecessary sub-sections
> >  - "dedicated mac" address no longer in current versions of document
> >  - "v4 in v6 / v6 in v4, etc." - intentionally unspecified since arbitrary
> >    encapsulations are supported by specification.  Implementations may have
> >    specific limitations.
> >  - "section 5 dedicated mac" no longer in the document
> >  - "decapsulation procedure reference" I believe has been clarified.)
> > 
> > (Mirja's comment on status is covered by Open Issue 2)
> > 
> > (Comments from Barry L. addressed:
> >  - "forming up"
> >  - bfd packet/vtep packets/vteps plurality agreement.
> >  - "may be configured" clarified.
> >  - Section 4.1 "of")
> > 
> > ---
> > 
> > Open Issue 25: "leading to a false negative" (COMMENT via Barry L.)
> > 
> > Proposed Action: The underlying concern in this sentence is that BFD packets
> > must not be mis-delivered to VMs since there will be no BFD machinery
> > present in that VM to execute the BFD procedures and thus sessions will
> > drop.  Possible action is to simply delete this sentence since it
> > prematurely anticipates procedures later described in the document.
> 
> This action is still pending.
> 
> > 
> > ---
> > 
> > Open Issue 26: "loopback range through a firewall" (COMMENT via Barry L.)
> > 
> > Proposed Action: Accept suggested rewording.
> 
> This action is still pending.
> 
> > ---
> > 
> > Open Issue 27: "Section 4...addresses the scenario" (COMMENT via Barry L.)
> > 
> > Proposed Action: This sentence needs to be reworded.
> 
> This issue is addressed by referring to GTSM mechanisms per other
> discussion.
> 
> > ---
> > 
> > (Comments from Adam R. addressed:
> >  - Form of ipv6 mapped address.
> >  - Usage of loopback network  addresses compared to prior RFCs discussed in
> >    thread with IESG.)
> > 
> > 
> > 
> > 
> > 
> > -- Jeff
> 
> >From jhaas as part of reviewing -12:
> 
> ---
> 
> Section 7 should be rewritten as:
> 7. BFD Echo Function
> 
> Support for the BFD Echo function [RFC5880], Section 3.2, is outside the
> scope of this document.
> 
> --- 
> 
> There remain a few minor English article agreement issues in the text, but
> these can be deferred to the RFC Editor's preference for resolution.
> 
> -- Jeff
> 
>