Skip to main content

Minutes IETF98: ospf
minutes-98-ospf-00

Meeting Minutes Open Shortest Path First IGP (ospf) WG
Date and time 2017-03-30 18:00
Title Minutes IETF98: ospf
State Active
Other versions plain text
Last updated 2017-04-03

minutes-98-ospf-00
IETF 98: Open Shortest Path First (OSPF) WG Agenda

Thursday, March 30th, 2017 (CDT) 13:00 - 15:00 - Afternoon Session I

Location: Zurich D
========================================================

Chairs: Acee Lindem
        Abhay Roy

WG Status Web Page: http://tools.ietf.org/wg/ospf/

Scribe: Les Ginsberg/Yingzhen Qu

1) Administrivia - 5 minutes
- Blue sheets
- Scribe/jabber
- Jabber room: ospf@jabber.ietf.org

2) WG Status Update - 10 minutes
- Acee Lindem

See slides.

Entropy Label: Defined more precisely in discussions w Stephane Litkowski.

Extensions for MRT:

Chris Bowers: Waiting on RTGWG draft revision - move a timer value out
of this document into the RTGWG draft. After that will update this draft.


3) OSPF Segment Routing Update - 10 Minutes
- https://datatracker.ietf.org/doc/draft-ietf-ospf-segment-routing-extensions/
- https://datatracker.ietf.org/doc/draft-ietf-ospf-ospfv3-segment-routing-extensions/
- Acee for Peter Psenak

See slides

4) OSPF YANG Model Update - 15 Minutes
- https://datatracker.ietf.org/doc/draft-ietf-ospf-yang/
- https://datatracker.ietf.org/doc/draft-ietf-ospf-sr-yang/
- Helen Chen

See slides

Rob Shakir: Updating policy model

Alia Atlas: Active discussion about combined config-oper or separate trees.
   What makes sense for the users?
   How do operators want to get state?
   Want an answer to that rather than what the policy is.

Chris Hopps: Does this augment routing model?

Helen Chen: Yes

Chris Hopps: Sad...

Acee Lindem: Could augment only one of the trees if data store was coming soon

Chris Hopps: Routing/routing state - put it under routing?

Acee Lindem: yes

Chris Hopps: Routing RFC specified - don't think this is still open.

Acee Lindem: Without policy model it is not useful. But that is a deployment concern.

Chris Hopps: Not blocking.

Jeff Haas: For BFD should use schema mount

Helen Chen: Will look at it

AceenLindem: Might want to put BFD state in oper state.
             IGPs only use BFD for down transition

Jeff Haas: IGPs have BFD timer config etc.

Acee Lindem: Association to an endpoint.
             Did use BFD grouping

Jeff Haas: How to stub out the feature - use schema mount in new draft

Chris Hopps: BFD session state is used to block adj up in IS-IS

Les Ginsberg: Some OSPF implementations also do this.

Chris Hopps: Datastore - Is there any conflict between config/oper
                         state 1:1.  Any system generated config?

Helen Chen: Some system generated oper state

Chris Hopps: Then why not use combined tree under routing - that is
             the recommendation.

Helen Chen: Thought that has changed

Acee Lindem: Discussed with Jeff Haas and others on BFD - config in protocol
             vs in BFD. Moved the config out of OSPF.

Yingzhen Qu: Back and forth on BFD parameters (protocol vs BFD)
             Now in BFD model.

5) OSPF TE Attributes for non-TE Applications Update - 10 Minutes
- https://datatracker.ietf.org/doc/draft-ppsenak-ospf-te-link-attr-reuse/
- Les Ginsberg

Chris Bowers: Author of the alternate draft. our draft focuses on fixing
              existing acknowledged problem. No change since Seoul, ready for
              last call, but there is no overall consensus in the WG.

Les Ginsberg: You said last call?

Chris Bowers: Sorry, WG adoption. we proposed a solution to move forward,
              looking at use cases for application value. I believe most use cases
              are vaguely described. so far the proposed draft doesn't
              justify what's being proposed. The drafts propose deprecating the
              existing advertisement.

Les Ginsberg: Last statement is not correct.

Chris Bowers: ISIS draft is explicitly deprecating the existing advertisement
              which are still useful. I believe the two drafts will add work
              for vendors and interop testing for vendors with no proven benefits.
              We should look at use cases and come up with a solution based on
              that instead of a  solution looking for use cases.

Les Ginsberg: We have deployment cases.  For example SR-TE and RSVP TE, but
              no way to differentiate which is which.  There might be use cases
              we don't know about now. We don't want to come back years later and
              find out this is not sufficient. We are building a framework that's
              flexible.

Chris Bowers: So far the uses case are not described. New requirements
              can be resolved without deprecating existing advertisements.

Les Ginsberg: OSPF draft doesn't propose deprecation. The IS-IS draft only
              suggest it is possible. We can resolve that easily.

Robert Raszuk: If I want to migrate from one to the other, how do I do it?

Chris Bowers: Use multi-topology routing.
              It would be useful to detail use cases. In general case, I'm not a
              big fan of a use case doc. but now we reach a situation where use
              cases are needed.

Acee Lindem: I'd like to see this draft moving forward. we'd prefer to have
             fewer LSAs to use. during migration, starting from using 3 LSAs to use
             the base LSA and extended prefix LSA. In ospfv3, only extended LSA is
             used.

Alia Atlas: It's a sensitive point as it is impacting existing deployments.
            People are making assumptions. I'm happy to talk offline with Acee
            and Abhay about it. if you have an implementation and are impacted by
            this, please email me. For operators, we'd like to see the value of this,
            and see what make sense going forward.

Acee Lindem: Both solutions have impacts on existing code.

Alia Atlas: Propose to have a virtual joint interim between OSPF and ISIS if
            needed. We want a similar solution solution and we want to hear
            impacts from both sides.  I fully understand someone will be unhappy,
            but we want to understand the deployment impact. I understand
            the pain.

Les Ginsberg: This is impacting deployments right now, because we don't have
              a standardized solution. Taking too long to get a standardized
              solution.

Alia Atlas: I get that. I was optimistic that we could put down use cases,
            weight pros and cons and make progress on this one. I'd work with
            all chairs to make progress, I want to get perspectives. I hear you on
            the configuration pain and I agree configuration is ugly, but...

Les Ginsberg: Even the authors of the alternate drafts agree their proposal is
              not good long term solution.

Alia Atlas: I brought it up because you said there is a problem and we don't
            have a solution. Sometimes you have to keep things running.

Les GInsberg: It's not a complete solution.

Alia Atlas: Yes, agreed.

Chris Hopps: I wasn't in Seoul. Did we talk about use case documents? We didn't
             get one. Before having an interim meeting, we need a use case
             document. You two can get the deprecation issue worked out.
             We need use cases doc before interim.

Les Ginsberg : We have at least one use case and a solution for application
               specific values.

Ahmed Bashandy: It seems the discussion is whether to get a long term solution
                or use configurations. We have a good solution.

Stephane Litkowski: I have one concern. if tomorrow we have a new application,
                    do we need to standardize a new bit?

Les Ginsberg: With both solutions, there is an extensible application bit mask.

Stephane Litkowski: When there is a new application, a new bit needs to be
                    standardized. Can you make it more generic? Support adding
                    ID dynamically?

Les Ginsberg: you are suggesting a tag?

Stephane Litkowski: a tag or something...

Les Ginsberg: Your suggestion is awkward from coding point of view.

Stephane Litkowski: Use different tag for different app.

Les Ginsberg: We're getting ahead of ourselves.

Stephane Litkowski: I don't want to stop at per application.

Les Ginsberg: I hear you on that. but there is a clear difference between
              the two solutions.

Stephane Litkowski: We want something generic, without the need to standardize
                    all applications.

Vishnu Pavan (TEAS chair): At TEAS WG, there is a topology document, and we
              can share it here.

Shraddha Hegde: One comment here. Per application attribute should not impact RSVP
                as nothing changed in RSVP.

Les Ginsberg: The OSPF draft doesn't suggest rsvp should be removed. ISIS draft
              only suggests it as a possibility, not a requirement.

Chris Bowers: Stephane's use case is more general and makes sense. You can
imagine in this proposal you're hard wiring a value for SR and there might be
two applications need two values.

Les Ginsberg: It's possible to share values between applications.

Chris Bowers: One controller doing TE, another doing LFA. Two SR applications
              to use different values. How do you resolve that?

Les: We support that.

Chris Bowers: I'd like to see that in use cases. I didn't see how that works.
              Can you work on a document to show that?

Les Ginsberg: I said I'd be happy to work on a use case document in Seoul.
              I didn't see it necessary, but I'm happy to work on the use
              case document.

Acee Lindem: We will look at some use cases and a possible interim scheduled.
            Thanks.

6) OSPF Flooding Reduction in MSDC - 10 Minutes
- https:https://datatracker.ietf.org/doc/draft-xu-ospf-flooding-reduction-in-msdc/
- Xiaohu Xu

See slides

Acee Lindem: Uses management LAN that connects to every switch in Data Center
             and floods all connectivity over that LAN.
             You should not advertise 2-way until FULL w DR.
             Analagous to route-reflector in BGP.


7) OSPF Extensions for Broadcast Inter-AS TE Link - 10 Minutes
- https:https://datatracker.ietf.org/doc/draft-chen-ospf-ias-lk/
- Huaimo Chen

See slides

Acee: Why don't you have prefix length?
       Not running OSPF on broadcast so it is a stub network.
       Minimize LSAs - so include prefix length in the encoding.

Huaimo: Will consider

Acee Lindem: How many think this is a valid use case? (2 hands)
             Not many people had read the draft

Acee Lindem: Want authors to initiate discussion


8) OSPF LLS Extensions for Local Interface ID Advertisement
- https: https://datatracker.ietf.org/doc/draft-ppsenak-ospf-lls-interface-id/
- Acee Lindem

See slides

Chris Bowers: Local IDs advertised in TE LSA

Les Ginsberg: How do you learn IDs if not running TE?

Chris Bowers: Can be solved - related to TE reuse discussion

Shraddha Hegde: What is the use case knowing this before adjacency is
established?

Acee Lindem: Don't like originating TE LSAs when you don't need to.

Tony Pryzienda: What happens when interface ID changes while adjacency is up?

Les Ginsberg: Should not have to use TE to get this functionality.
              IS-IS proves that it works.

Acee Lindem: Do you advertise a TE LSA on every link?

Chris Bowers: When we need to.

9) OSPF Extensions for Advertising/Signaling BGP Route Reflector Information - 10 Minutes
- https:https://datatracker.ietf.org/doc/draft-acee-ospf-bgp-rr/
- Acee Lindem


See slides

No comments