Re: [Idr] WG Adoption for draft-zhu-idr-bgp-ls-path-mtu (11/1/2020 to 11/16/2020)

["Pengshuping (Peng Shuping)" <[e-mail]>]

Hi Sue,

Yes, it can exist in a BGP-only solution. In fact there are many ways to obtain the link MTU information, and it does not have to be from the IGP LSDB.

An example topology for the BGP-only case could be a cross-domain tunnel, i.e. a tunnel over a few domains. In order to compute the Path MTU of this cross-domain tunnel, it needs to get the tunnel MTU of the sub-tunnel in each domain, which could be directly delivered to the controller from the ingress node of each domain using BGP-LS.

Best regards,
Shuping


From: Idr [mailto:[e-mail]] On Behalf Of Susan Hares
Sent: Wednesday, November 11, 2020 5:11 AM
To: 'Acee Lindem (acee)' <[e-mail]>; 'Stephane Litkowski (slitkows)' <slitkows=[e-mail]>; [e-mail]
Subject: Re: [Idr] WG Adoption for draft-zhu-idr-bgp-ls-path-mtu (11/1/2020 to 11/16/2020)

Acee:

I agree with you that the name of the draft is wrong.    I hope the authors will resubmit the draft with a revised name before the LSR /IDR sessions.

If LSR wants to pick up the draft as there is no link MTU, this addresses Stephane’s concern.  The one concern I need Shuping to answer is whether this function can exist in a BGP-only solution.

If can be in BGP-only LSRs then we need to make sure the BGP-only solution works.   If it goes to LSR, we’ll need cross review.   Since LSR is before IDR on Monday, we’ll let this WG adoption call continue to get more comments.

During IETF the IDR and LSR chairs can agree where the document goes.

Cheers, Sue

From: Acee Lindem (acee) [mailto:[e-mail]]
Sent: Tuesday, November 10, 2020 4:01 PM
To: Susan Hares; 'Stephane Litkowski (slitkows)'; [e-mail]<mailto:[e-mail]>
Subject: Re: [Idr] WG Adoption for draft-zhu-idr-bgp-ls-path-mtu (11/1/2020 to 11/16/2020)

Speaking as an IDR WG member:

The name of the draft is wrong – the extension is for a Link MTU and not a path MTU.

Speaking as LSR Chair:

We could this in LSR as there is currently no MTU advertisement in the LSAs for OSPFv2 and OSPFv3. Implementations already make use of this information as it is used in the OSPF DBD packets and for LSA packing. Of course, we’d require a more accurate draft name and title.

Thanks,
Acee

From: Idr <[e-mail]<mailto:[e-mail]>> on behalf of Susan Hares <[e-mail]<mailto:[e-mail]>>
Date: Monday, November 9, 2020 at 4:20 PM
To: "'Stephane Litkowski (slitkows)'" <slitkows=[e-mail]<mailto:slitkows=[e-mail]>>, IDR List <[e-mail]<mailto:[e-mail]>>
Subject: Re: [Idr] WG Adoption for draft-zhu-idr-bgp-ls-path-mtu (11/1/2020 to 11/16/2020)

Stephane:

My second message to this thread asked a few questions about the technology.

This information can be more than IGP information.   If SR segments statically defined (static or direct interfaces) tunnels and pass the endpoints via BGP tunnel-encaps draft with SR Policy tunnel type, this can just be BGP.

I’ll keep this WG adoption call going until we can be sure if:  1) it something LSR wants to standardize, and 2) whether there is a BGP only case.   It is clear to me that standardizing MTU for a SR segments with stacked tunnel segments passed by BGP was useful.

The authors should be the ones to propose this in LSR.

Cheers,  Sue

From: Idr [mailto:[e-mail]] On Behalf Of Stephane Litkowski (slitkows)
Sent: Monday, November 9, 2020 4:28 AM
To: Susan Hares; [e-mail]<mailto:[e-mail]>
Subject: Re: [Idr] WG Adoption for draft-zhu-idr-bgp-ls-path-mtu (11/1/2020 to 11/16/2020)

Hi Sue,

> The purpose behind this mechanism is to reduce administrative work rather than to reduce the review on drafts.

That’s exactly my point. If we don’t do OSPF extension now and in the same draft, we leave a gap that will require a new draft for a very very small extension. Just adds process overhead for nothing…


Stephane


From: Susan Hares <[e-mail]<mailto:[e-mail]>>
Sent: lundi 9 novembre 2020 10:10
To: Stephane Litkowski (slitkows) <[e-mail]<mailto:[e-mail]>>; [e-mail]<mailto:[e-mail]>
Subject: RE: [Idr] WG Adoption for draft-zhu-idr-bgp-ls-path-mtu (11/1/2020 to 11/16/2020)

Stephane:

I want to pick up on your email from two points:

1)  Why not do everything in LSR?
<WG-chair hat>
If the feature comes with interest in doing all 3 (ISIS, OSPF, and BGP-LS data gathering), then the authors may select to do everything in LSR rather than have 2 or 3 drafts to maintain.

This is optional and the mechanism may not fit every draft.   The drafts may also start out adopted and vetted in LSR and IDR.    The purpose behind this mechanism is to reduce administrative work rather than to reduce the review on drafts.

</wg-chair hat off>


2) TRILL implementations of IS-IS has some MTU subTLV -

If you are interested in whether this has been implemented in TRILL, you might want to check with Donald Eastlake.   My vague and foggy recollection is that had some implementations or came from pre-TRILL implementations.


Cheers, Susan Hares



From: Stephane Litkowski (slitkows) [mailto:[e-mail]]
Sent: Wednesday, November 4, 2020 3:03 AM
To: Susan Hares; [e-mail]<mailto:[e-mail]>
Subject: RE: [Idr] WG Adoption for draft-zhu-idr-bgp-ls-path-mtu (11/1/2020 to 11/16/2020)

Hi,

“a) Are there ways to pass IGP link MTUs in
the IGPs?  If so, is this needed in BGP-LS”

This is a valid point, most of the time BGP-LS is feeded by IGP LSDBs (of course there are other ways too). While I see that IS-IS has some MTU subTLV coming from TRILL RFC7176 (possibly never been implemented), I don’t see anything for OSPF (I’m not an OSPF expert, so I may have missed it).
Shouldn’t this be checked and validated with LSR WG before adopting ?


Stephane


From: Idr <[e-mail]<mailto:[e-mail]>> On Behalf Of Susan Hares
Sent: lundi 2 novembre 2020 06:04
To: [e-mail]<mailto:[e-mail]>
Subject: [Idr] WG Adoption for draft-zhu-idr-bgp-ls-path-mtu (11/1/2020 to 11/16/2020)

This begins a 2 week WG adoption call for
draft-zhu-idr-bgp-ls-path-mtu-04.txt (11/1 – 11/16/2020).

The authors should send in an IPR statement for this draft
by 11/5 so the WG can include the IPR status in their decision.

You can access the draft at:
https://datatracker.ietf.org/doc/draft-zhu-idr-bgp-ls-path-mtu/

Since this draft is reference by an existing IDR draft
I’ve included a bit of background below to help you place
this draft into the larger context of the SR additions to BGP-LS
and the draft-ietf-idr-tunnel-encaps-19.txt.

This draft does continue BGP-LS additions.  if you
are opposed to any BGP-LS additions rather than
this specific addition, please make that clear in your
comment in this discussion.

The authors requested a WG adoption at IETF 108.
The IDR co-chairs thank the authors for their patience.
This draft has been delayed by process of having a
new document shepherd (Sue Hares) come up to speed
on draft-ietf-idr-tunnel-encapsulation.

Cheers, Sue

Background
===========
Segment Routing technology creates SR tunnels that are
directly overlaid on MPLS or SRv6.  While existing MPLS technology
(LDP and RSV-TE) provides mechanisms to negotiate path MTU
based on individual link MTU limits, the Segment Routing (SR)
on BGP-LS Link Attribute does not pass information on
MTU size per link.

draft-ietf-idr-sr-policy-path-mtu-02.txt sends PATH MTU
information in the tunnel-encapsulation attribute for the tunnel type
SR-Policy that handles segment routing (SR) paths.
However, it lacks the information to create a reasonable
Path size since the BGP-LS Link Attribute does distribute
this information.

The draft proposes adding a new sub-TLV for MTU size
to the BGP-LS Link Attribute TLV, and
draft-ietf-idr-sr-policy-path-mtu-02.txt mentions this
draft as one possible way to distribute the per link
MTU.

Questions for the authors might be:
a) Are there ways to pass IGP link MTUs in
the IGPs?  If so, is this needed in BGP-LS

b) What other mechanisms pass link MTU?