Re: [Idr] Growing BGP-LS Attribute

[Zhuangshunwan <[e-mail]>]

From: Idr [mailto:[e-mail]] On Behalf Of Robert Raszuk
Sent: Monday, October 22, 2018 3:45 AM
To: [e-mail]
Cc: [e-mail]
Subject: Re: [Idr] Growing BGP-LS Attribute


> soon we will have bgp-ls doctors.

Honestly having to read through all of the recent bgp-ls proposals and extensions there is so much confusion there on what goes to NLRI vs BGP-LS Attribute, what is TLV and when nesting to sub-TLVs makes sense, why do we send different BGP-LS Attribute depending to which NLRI it accompanies etc .. that I think BGP-LS doctor(s) are needed not soon but yesterday. And some of those documents are at the the WG-LC status too ...

Yes interestingly no one cares to touch on few other BGP aspects which may come into play. Example: Should BGP-LS be eligible to be packed into BMP (considering recent BMP extension proposals), how the dynamic IGP data be of any use if some BGP still uses long MRAI timer for subsequent updates on the same NLRI etc ...
[Shunwan]
I think the idea of using BMP is feasible.
The BMP Loc-RIB draft (https://tools.ietf.org/wg/grow/draft-ietf-grow-bmp-local-rib/) already mentions this if I understand it correctly.
It says:
"
   o  Requires additional BGP peering to collect the received updates
      when peering may have not even been required in the first place.
      For example, VRF's with no peers, redistributed bgp-ls with no
      peers, segment routing egress peer engineering where no peers have
      link-state address family enabled.

"
Thanks,
Shunwan

> as adrian says, if you wanna monitor the igp, then do so directly.

Well as you know in practice this is not always an option - in fact this is almost never an option in production network ... Controllers running on x86 boxes sit far from routers and are never allowed to connect to them directly. So bringing IGP adjacency between router and such oracle is hard ... even assuming it would be just passive one. That's why TCP is so attractive since as long as server is reachable the session will be formed and bits transferred..


On Sun, Oct 21, 2018 at 8:33 PM Randy Bush <[e-mail]<mailto:[e-mail]>> wrote:
> I think by all means this is the largest single BGP attribute we ever
> had and it seems growing daily like a balloon. Most of the TLVs which
> are defined contain sub-TLVs which can be filled with data in a non
> limited by any spec fashion.

this is scary.  as adrian says, if you wanna monitor the igp, then do so
directly.  this bgp-ls thing is trying to funnel all of bgp and igps;
but to no clear end.  and, as igps and bgp keep expanding, bgp-ls will
only amplify the problem.  soon we will have bgp-ls doctors.

and i wonder why it does not also funnel dns. :)

randy