Re: [Idr] BGP MIBv2 implementation

[Bill Fenner <[e-mail]>]

On Sun, Jul 22, 2012 at 6:53 AM, Daniel Cohn <[e-mail]> wrote:

> Hi Bill and all, please see inline with [DC].****
>
> ** **
>
> *From:* [e-mail] [mailto:[e-mail]] *On Behalf Of *Bill
> Fenner
> *Sent:* Tuesday, July 17, 2012 9:59 PM
> *To:* Daniel Cohn
> *Cc:* Jeffrey Haas; [e-mail]
> *Subject:* Re: [Idr] BGP MIBv2 implementation****
>
> ** **
>
> On Wed, Apr 25, 2012 at 8:44 AM, Daniel Cohn <[e-mail]> wrote:**
> **
>
> Hi Jeff and all,****
>
>  ****
>
> Sorry for the delayed response. I think it would be a shame to go through
> the hassle of updating the BGP MIB without fully supporting BGP L3VPN/VPLS
> (in both RFC 4761 and 6074 flavors). Therefore I support using a generic
> NLRI prefix TC that explicitly supports these NLRIs and can be extended to
> support future BGP applications. It doesn’t seem consistent to add AFI/SAFI
> to the MIB while leaving the NLRI as an IP address.****
>
>  ****
>
> A couple more suggestions for this draft:****
>
>  ****
>
> -          Change the bgp4V2AdjRibsOutTable index to make it the same as
> in bgp4V2NlriTable. In its current form (a pointer to the bgp4V2NlriTable),
> it doesn’t support BGP routes originated by this speaker (e.g.
> redistributed from IGP) so it’s woefully inadequate.****
>
> I had a similar thought about the Adj-RIBS-Out table - it can't represent
> a locally-originated route.  I am not sure what you mean by changing the
> index to address this, however, since the index as it is makes sense to me.
> ****
>
> ** **
>
>  Maybe you mean to change the structure of the table to copy all of the
> data from the NlriTable, which I disagree with.****
>
> ** **
>
> [DC] You’re right, that’s what I meant. My bad. ****
>
> ** **
>
> One idea on handling this with the current structure is to use the pointer
> to point to a row in the IP-FORWARDING-MIB for a locally-originated route.
> ****
>
> ** **
>
> [DC] I don’t think that would work, as there is plenty of BGP-specific
> information that is not available at the IP forwarding MIB, such as MED,
> LOCAL_PREF, or ext communities (actually all the information currently
> indexed by bgpM2PathAttrIndex).****
>
> What is the problem with expanding the table to include all required BGP
> information? If your concern is memory utilization, at the risk of stating
> the obvious let me say that MIB structure doesn’t necessarily prescribe the
> internal database structure, meaning you can always choose a more efficient
> internal structure (e.g. using pointers instead of storing duplicate
> information) and translate the SNMP queries from/to the MIB structure.
>

Of course the SNMP structure is not necessarily at all related to the
internal data structures.  My understanding is that the reason for the
Rib-OUT table using RowPointer was the [not that well-founded, but still
widespread] concern about the amount of data returned in a full MIB walk.
 (Also, it does concretely reduce the amount of data you need to process if
you're a management system that wants to know all of the RIB-IN plus
RIB-OUTs.)

What about representing locally-advertised routes as being received from a
peer of type unknown, with zero length address?


> ****
>
> -          Add support for BGP communities (as in your communities draft)
> in the NLRI in/out tables.****
>
> There is no reason this can't be a future addition, right, in the interest
> of getting the base functionality out?****
>
> ** **
>
> [DC] Why should this simple addition hold back progress?
>

It was removed in 2007 because of a lack of consensus about including it.
 Is your sense that that lack of consensus has changed?


> -          Restore local address (and maybe ports) to bgp4V2PeerTable
> index to support multiple BGP sessions between two speakers as in
> draft-ietf-grow-diverse-bgp-path-dist****
>
> If the ports are in the INDEX, how do you ever find a row if one of the
> ports is guaranteed to be ephemeral and you don't know a priori who
> originated the connection?****
>
> ** **
>
> If the remote address was first in the INDEX, you could always use getnext
> to get a given value independent of the local address, but that's not a
> concept that many management systems are familiar with.****
>
> ** **
>
> An idea that solves both the ports and the unknown-local-address problem
> is to use the remote address and an instance identifier.  Peers with a
> single connection (the "normal" case) would be only instance 1, but if you
> had multiple simultaneous connections to the same remote peer you could use
> more values.****
>
> ** **
>
> [DC] I don’t see a problem with using getnext, but a neutral instance is
> also acceptable.
>

The problem with using getnext is that so many management systems are
"MRTG-like" in that they can graph arbitrary objects if you give them the
whole OID, but can only use GET to do so.  You may say that people
shouldn't use such management systems, but if there's a straightforward way
to accomodate them then why not?

  Bill

 ****
>
> Feedback is welcome, regards,****
>
>  ****
>
> Daniel****
>
>  ****
>
> *From:* Jeffrey Haas [mailto:[e-mail]]
> *Sent:* Thu, 15 Mar 2012 14:19:17
> *Subject:* Re: [Idr] IETF 83 - BGP MIBv2 Update - prefix Textual
> Conventions****
>
>  ****
>
> Working Group,****
>
>  ****
>
> As per prior plenary discussion on making more effect use of working group
> ****
>
> session time, please find my BGP MIBv2 update slides on the proceedings***
> *
>
> page:****
>
>  ****
>
> http://www.ietf.org/proceedings/83/slides/slides-83-idr-0.pdf****
>
>  ****
>
> The bulk of the presentation is self explanatory.  I'd like to thank ****
>
> Bert Wijnen for additional MIB feedback.   I suspect we'll get a bit more*
> ***
>
> feedback from him since he discovered bugs as part of doing derivative work
> ****
>
> for the SIDR MIB.  Draft -13 [1] reflects his comments as of the IETF 83**
> **
>
> posting cutoff.****
>
>  ****
>
> One item I would like to draw the working group's further attention to is*
> ***
>
> covered on slides 5-7.  As part of analyzing the requirements for the****
>
> multicast next-gen VPN [2] MIBs that Jeffrey Zhang (zzhang at juniper.net)
> is****
>
> working on, we briefly considered exposing MVPN routing state in BGP via
> the****
>
> BGP MIBv2.  Per prior presentations and discussions on design goals for the
> ****
>
> BGP MIBv2, we wanted one MIB that was capable of being re-used for various
> ****
>
> types of reachability that was being carried in BGP.  However, the primary
> ****
>
> goal was to be able to carry IPv6 reachability.****
>
>  ****
>
> RFC 6514 can be briefly examined to see the different types of reachability
> ****
>
> that are being passed around in BGP.  The reachability is distinguished***
> *
>
> within the same AFI/SAFI by a "Route Type" field within the prefix with
> type****
>
> specific encodings following.  ****
>
>  ****
>
> The BGP MIBv2 presents prefixes using the following two objects:****
>
>     bgp4V2NlriPrefixType OBJECT-TYPE****
>
>         SYNTAX     InetAddressType****
>
>         MAX-ACCESS not-accessible****
>
>         STATUS     current****
>
>         DESCRIPTION****
>
>             "The type of the IP address prefix in the****
>
>              Network Layer Reachability Information field.****
>
>              The value of this object is derived from the****
>
>              appropriate value from the bgp4V2NlriAfi field.****
>
>              Where an appropriate InetAddressType is not****
>
>              available, the value of the object must be****
>
>              unknown(0)."****
>
>         ::= { bgp4V2NlriEntry 4 }****
>
>  ****
>
>     bgp4V2NlriPrefix OBJECT-TYPE****
>
>         SYNTAX     InetAddress****
>
>         MAX-ACCESS not-accessible****
>
>         STATUS     current****
>
>         DESCRIPTION****
>
>             "An IP address prefix in the Network Layer****
>
>              Reachability Information field. This object****
>
>              is an IP address containing the prefix with****
>
>              length specified by bgp4V2NlriPrefixLen.****
>
>              Any bits beyond the length specified by****
>
>              bgp4V2NlriPrefixLen are zeroed.****
>
>  ****
>
>              An implementation is required to support IPv4****
>
>              prefixes.  In this case, the object length****
>
>              is (0..4).****
>
>  ****
>
>              An implementation MAY support IPv6 prefixes.****
>
>              In this case, the object length is (0..16)"****
>
>         REFERENCE****
>
>             "RFC 4271, Section 4.3."****
>
>         ::= { bgp4V2NlriEntry 5 }****
>
>  ****
>
> The general goal of the "InetAddressType/InetAddress" textual conventions
> [3]****
>
> is to provide a level of generality for MIB objects that are "addresses".*
> ***
>
> Our initial design guidance for the BGP MIBv2 was to make use of these TCs
> ****
>
> to help represent both IPv4 and IPv6 addresses - and it does this well.***
> *
>
>  ****
>
> As part of investigating encoding MVPN prefixes in the BGP MIBv2, I sent**
> **
>
> mail to the ietfmibs mailing list to discuss this possibility.  After a few
> ****
>
> exchanges, it was suggested that since the information is BGP specific that
> ****
>
> this was not the best fit for the INET-ADDRESS-MIB and that we should****
>
> consider a BGP TC specifically for this purpose.****
>
>  ****
>
> If we were to consider such a TC, we would make it "IANA maintained".  This
> ****
>
> would permit us to issue the initial MIB but permit IANA to maintain the**
> **
>
> code points as new protocol elements are added.  I.e. we don't have to have
> ****
>
> a MIB document stuck as an I-D forever.  The current INET-ADDRESS-MIB is**
> **
>
> an example of such a MIB.****
>
>  ****
>
> After further discussion with Jeffrey Zhang, we determined that there
> wasn't****
>
> a strong incentive to continue the MVPN MIB work as a BGP MIBv2 extension,
> ****
>
> We may see future requirements from other BGP-based address families.  ***
> *
>
>  ****
>
> My recommendation would be that we proceed with the work to create a BGP**
> **
>
> Prefix TC MIB with an explicit goal of being backward compatible with the*
> ***
>
> INET-ADDRESS-MIB.  However, the WG may also come to the decision to not***
> *
>
> pursue making this MIB completely general purpose and just worry about****
>
> IPv4/IPv6.****
>
>  ****
>
> With regard to standards advancement, the new BGP Prefix MIB shouldn't****
>
> unnecessarily hinder the BGP MIBv2.****
>
>  ****
>
> I would like the WG to come to some consensus as to whether we (very
> likely,****
>
> I) take on the work to make such a prefix TC MIB.****
>
>  ****
>
> If the discussion becomes interesting enough, Sue and John have agreed to*
> ***
>
> grant us discussion time at the upcoming IDR session in Paris.****
>
>  ****
>
> -- Jeff****
>
>  ****
>
>
> _______________________________________________
> Idr mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/idr****
>
> ** **
>