Re: [spring] draft-ginsberg-spring-conflict-resolution - WG adoption call

["Stefano Previdi (sprevidi)" <[e-mail]>]

> On May 6, 2016, at 10:16 PM, Uma Chunduri <[e-mail]> wrote:
> 
> Les,
>  
> 2 quick things.
>  
> 1.        
>             >[Les:] There are two legitimate use cases for SRMS:
>                                            >1)To advertise SIDs for non-SR capable nodes
>                                            >2)As a global provisioning tool
>                          I am hearing #2 for the first time. I don’t see this either  discussed earlier in the WG list  or captured in architecture document 
>                          https://tools.ietf.org/html/draft-ietf-spring-segment-routing-07. Even in the protocol extensions document for example 
>                          https://tools.ietf.org/html/draft-ietf-isis-segment-routing-extensions-06#section-2.4 we always discussed this from non-SR 
>                          capable nodes PoV. So I request to add this in architecture document before factoring this for solution in conflict resolution. 


using a SRMS for advertising SID on behalf of SR capable nodes does not introduce any change in the SR architecture so not sure what we need to document here.


                       
> 
> 2.       Also this is the first time ever we have a requirement for cross protocols verification we ought to discuss the implications of this  
> and protocols involved (with in AS or otherwise) in the architecture document (at least briefly).


the architecture draft is data-pane agnostic so I presume you refer to draft-ietf-spring-segment-routing-mpls.

with the ipv6 data-plane SR conflicts are in fact solved by ipv6 addressing techniques ;-)

s.


>  
> --
> Uma C.
>  
> From: spring [mailto:[e-mail]] On Behalf Of Les Ginsberg (ginsberg)
> Sent: Wednesday, May 04, 2016 9:38 AM
> To: Uma Chunduri; [e-mail]; [e-mail]
> Subject: Re: [spring] draft-ginsberg-spring-conflict-resolution - WG adoption call
>  
> Uma –
>  
> To restate, the problem being addressed here is to guarantee consistent use of SIDs in the forwarding plane network-wide in the presence of conflicting advertisements. The set of advertisements includes both SIDs advertised in protocol prefix reachability advertisements and SRMS advertisements because problems occur based upon inconsistencies in what is installed in the forwarding plane of different routers. It matters not whether Router A used a SID advertised by a protocol prefix reachability advertisement or by an SRMS advertisement – what matter is whether the SID used is consistent with what the neighbors of Router A use. So simply ensuring that OSPF (for example) resolves SIDs in a consistent way does not fully address the problem space.
>  
> More inline.
>  
> From: Uma Chunduri [mailto:[e-mail]] 
> Sent: Tuesday, May 03, 2016 3:59 PM
> To: Les Ginsberg (ginsberg); [e-mail]; [e-mail]
> Subject: RE: [spring] draft-ginsberg-spring-conflict-resolution - WG adoption call
>  
> Les,
>  
> With all due respects, cross protocol verification  of prefix and SID conflicts as an “architectural change”  and it can severely impact the existing implementations (at least the one I work on). 
>  
> [Les:] It is quite correct – and I can confirm based on personal experience – that support for conflict resolution is a significant effort.
>  
> Also I have couple of cases where current version of the draft is not clear about resolution.
>  
> IMO, first we need clarity with in a protocol instance resolution rules before going to resolve the same across protocols (I mentioned few cases below) .
> Separation of reachability advertisements and SRMS would help “cross protocol” verification of the ranges and SRMS is not applicable to all protocols.
>  
>  
> In-line [Uma]:
>  
> --
> Uma C.
>  
> From: Les Ginsberg (ginsberg) [mailto:[e-mail]] 
> Sent: Saturday, April 30, 2016 10:11 PM
> To: Uma Chunduri; [e-mail]; [e-mail]
> Subject: RE: [spring] draft-ginsberg-spring-conflict-resolution - WG adoption call
>  
> Uma –
>  
> We are indeed defining conflict resolution across all the SID advertisements regardless of source (protocol or SRMS) –
>  
> [Uma]: While you can theoretically do anything for current implementation this kind of cross protocol verification is a new architectural requirement.  
>                Because it seems “a central entity” need to gather all different protocol instances SRMS advertisements and should settle with resolution. 
>  
> -          Also note SRMS is not applicable for BGP but it seems all prefix SIDs need to be verified  with IGP instances SRMS advertisements. Is this true? While the document mostly talks about these and compares with prefix advertisement.
> [Les:] The issue is protocol agnostic.
>  
> -          Algorithm proposed needs more clarity: Take Section 3.2.4
>  
> o    
>                       “   1.  Smaller range wins
>  
>              2.  IPv6 entry wins over IPv4 entry
>              …
>         “
>                  What happens in case of prefix conflict or SID conflict with only prefix advertisements (range 1).  Say multiple prefixes have same SID in one protocol instance and in different protocols.
>                  I see 2 levels of resolution required viz., one at within the protocol and one among the protocols.  No discussion on this.
>  
> [Les:] The full set of rules specified in the draft provide deterministic resolution in all cases. You have snipped only the first two rules. If a preference is not obtained based on the first two rules you continue on to rule #3, then rule #4, etc.
>  
>                  Similarly in case of SID conflict  (range 1), it’s not specified which protocol’s SID need to be considered.  Are you assuming some sort of Admin distance play a role in resolution? 
> [Les:] No – admin distance plays no role here.
>  
>  I don’t see any discussion in the document  and needs more clarity there too.
> o   Also what happens if a prefix or SID conflict happens with SRMS range 1 and a prefix  advertisement (2 cases)
> a.       of one protocol and
> b.      multiple protocols?
>  
> [Les:] The source of the SID advertisement (what protocol/protocol instance or whether it is SRMS based) plays no role. The tie breaker rules as defined are complete and provide a deterministic answer in all cases.
> If you believe that is not true please provide a specific example where you apply all the rules in the order specified and still do not determine the preferred entry.
>  
>  
> -          On the below assumption: (Section 3.2.4)
>                                          “This has the nice property that a single misconfiguratoion of an SRMS
>                  entry with a large range will not be preferred over a large number of
>                  SIDs advertised in prefix reachability advertisements.”
> While anything can happen in theory, I found it bit odd to see why SRMS entry is being advertised and for the same prefix, SID is also advertised through reachability advertisements? 
> This is against the spirit of SRMS advertisement, isn’t it? While this can happen, it seems we are claiming resolution solution by focusing more on  this case in the current version of the document.
>  
> [Les:] There are two legitimate use cases for SRMS:
>  
> 1)To advertise SIDs for non-SR capable nodes
> 2)As a global provisioning tool
>  
> Let’s examine the first case. I have an LDP enabled network and I begin introducing SR capable nodes. At a given moment in time Router A is NOT SR capable and SRMS advertisements cover prefix SIDs for the addresses associated with Router A.
> I then upgrade Router A to become SR capable. Because I want to do “make-before-break” I do not immediately remove the SRMS advertisements covering the addresses associated with Router A. I upgrade A, add configuration of SIDs locally on Router A, and verify that the advertisements originating from protocols on Router A are correct. If an inconsistency is introduced when configuring the SIDs on Router A then I will have an SRMS advertisement and a prefix-reachability advertisement that conflict. Until the conflict is corrected we use the conflict resolution rules to provide deterministic forwarding behavior.
>  
> This to me is a normal and expected upgrade scenario.
>  
>  
> This is one of the reasons of my first comment below. You got to separate the reachability advertisements and SRMS advertisements; as in principle these are defined for different purposes. I don’t see we  need algorithm to prefer reachability advertisement over SRMS advertisement (if we don’t compare these 2 categories).
>  
>  
>  
> [Les:] I disagree – hopefully my comments have helped you to understand why.
>  
>    Les
>  
>  
> as the sections you have quoted clearly state.
>  
> Why? Because we need consistent use of SIDs in the forwarding plane. From forwarding perspective it matters not whether the SID was advertised by protocol instance #1 or #2 or by an SRMS. 
>  
> What matters is that the SID I use to determine what label I install in my forwarding plane is the same SID that my neighbors will use. Otherwise forwarding will be broken.
>  
>    Les
>  
>  
> From: spring [mailto:[e-mail]] On Behalf Of Uma Chunduri
> Sent: Wednesday, April 27, 2016 4:31 PM
> To: [e-mail]; [e-mail]
> Subject: Re: [spring] draft-ginsberg-spring-conflict-resolution - WG adoption call
>  
> Dear Authors,
>  
> Have few comments on the draft:
>  
> 1. 
>         As I indicated during meeting - I am not sure why we have to club  verification of SIDs advertised through regular protocol reachability
>                 prefixes and the ranges advertised through 'Mapping Server'  (SRMS). I didn't see any compelling reason to do this.
>                  SIDs advertised through reachability prefixes doesn't have ranges unlike SRMS advertisements. 
>                  As SRMS advertisements are primarily for nodes which are not SR capable and  I feel we should not mix this with nodes which are SR capable.
>  
>         This isolation helps restricting the resolution work primarily for multiple SRMS entries advertised through one node or multiple nodes.
>                 SRMS advertisements are indeed little bit unique in that you are advertising "configuration" on behalf of node X from node Y
>                 with ranges (both prefix ranges and SID ranges).
>  
>  
> 2. 
>                 Regarding  the scope of conflict resolution:
>  
>  
>        Section 1
>  
>            "   The problem to be addressed is protocol independent i.e., segment
>          related advertisements may be originated by multiple nodes using
>          different protocols and yet the conflict resolution MUST be the same
>          on all nodes regardless of the protocol used to transport the
>          advertisements."
>  
>         Section 3.2.8
>           "   o  In cases where multiple routing protocols are in use mapping
>       entries advertised by all routing protocols MUST be included."
>  
>       This sounds like we are seeking to resolve conflicting entries outside and across the protocols?
>       Each IGP has separate mechanism for advertising mapping entry and I this is not clear with the current version of the draft how we can cross verify SID/Prefix conflict across  the protocol.
>      Can you clarify this?
>  
>  
> --
> Uma C.
>  
>  
> -----Original Message-----
> From: spring [mailto:[e-mail]] On Behalf Of [e-mail]
> Sent: Thursday, April 14, 2016 12:55 AM
> To: [e-mail]
> Subject: Re: [spring] draft-ginsberg-spring-conflict-resolution - WG adoption call
>  
> > From:  [e-mail] > Sent: Thursday, April 14, 2016 9:51
> > AM
> >
> > Dear WG,
> >
> > As we discussed at our meeting last week, working group adoption has
> > been requested for draft-ginsberg-spring-conflict-resolution.
> > Please reply to the list with your comments, including although not
> > limited to whether or not you support adoption.
>  
> We will end the call on April 29, 2016.
>  
>  
> > Thanks,
> >
> > --John and Bruno
> >
> >
> >
> > __________________________________________________________
> > __________________________________________________________
> > _____
> >
> > Ce message et ses pieces jointes peuvent contenir des informations
> > confidentielles ou privilegiees et ne doivent donc pas etre diffuses,
> > exploites ou copies sans autorisation. Si vous avez recu ce message
> > par erreur, veuillez le signaler a l'expediteur et le detruire ainsi
> > que les pieces jointes. Les messages electroniques etant susceptibles
> > d'alteration, Orange decline toute responsabilite si ce message a ete altere, deforme ou falsifie. Merci.
> >
> > This message and its attachments may contain confidential or
> > privileged information that may be protected by law; they should not
> > be distributed, used or copied without authorisation.
> > If you have received this email in error, please notify the sender and
> > delete this message and its attachments.
> > As emails may be altered, Orange is not liable for messages that have
> > been modified, changed or falsified.
> > Thank you.
> >
> > _______________________________________________
> > spring mailing list
> > [e-mail]
> > https://www.ietf.org/mailman/listinfo/spring
>  
> _________________________________________________________________________________________________________________________
>  
> Ce message et ses pieces jointes peuvent contenir des informations confidentielles ou privilegiees et ne doivent donc pas etre diffuses, exploites ou copies sans autorisation. Si vous avez recu ce message par erreur, veuillez le signaler a l'expediteur et le detruire ainsi que les pieces jointes. Les messages electroniques etant susceptibles d'alteration, Orange decline toute responsabilite si ce message a ete altere, deforme ou falsifie. Merci.
>  
> This message and its attachments may contain confidential or privileged information that may be protected by law; they should not be distributed, used or copied without authorisation.
> If you have received this email in error, please notify the sender and delete this message and its attachments.
> As emails may be altered, Orange is not liable for messages that have been modified, changed or falsified.
> Thank you.
>  
> _______________________________________________
> spring mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/spring
>  
> _______________________________________________
> spring mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/spring