Re: [GROW] BGP Auto Discovery and BGP Auto Session Setup

[Alejandro Acosta <[e-mail]>]

Hello Robert,

 I just read the first document, I guess for most people could be
"interesting" too see BGP Auto Discovery, I'm not sure yet if I'm in
favor or against.

 Hope this is not a crazy comment. Well, in the document you (and other
co-authors) do not mention much about -any possible- impact this
mechanism could have in bgp route selection algorithm (which anyhow is
by default managed by vendors and modified by administrators). As far as
I could understand from your doc, routers can identify which prefixes
were learnt by auto-discovered devices and which were learnt by the
"traditional" BGP operation. Am I right? which one would you trust
more?, new administrative distance or something like that?.


Thanks,


Alejandro,


On 12/11/19 3:24 PM, Robert Raszuk wrote:
> Dear WG,
>
> We have seen formation of BGP Autodiscovery WG followed by total
> silence. Well maybe group is 
> live just not everyone got onto its private list :) 
>
> Regardless of this I refreshed and resubmitted two proposals in this
> very space. 
>
>
>   1. BGP Auto Discovery
>
>
>   https://tools.ietf.org/html/draft-raszuk-idr-bgp-auto-discovery-06
>
>
>   and
>
>
>   2. BGP Automated Session Setup
>
>
>   https://tools.ietf.org/html/draft-raszuk-idr-bgp-auto-session-setup-01
>
> *_Ad 1:_*
> First document is focused on WAN and IX scenario where establishing
> full mesh of IBGP or
> selected eBGP peering can automate the operational management tasks or
> if run in informational
> mode could validate NMS session setup correctness.
> Original proposal was presented many years ago in Vienna and at that
> point suggested extension to
> IGPs to flood discovery information for IBGP auto mesh. Later based on
> the input and discussions with
> Pedro the proposal got simplified to use classic Route Reflector as
> bootstrap node (info broker).
> Last input from Jon and Warren added the ability to also automate
> setup of eBGP sessions in selected
> scenarios.
> Since then the document was shelved for some time waiting for IDR WG
> turn to deal with discovery topic.
> So here we are.
> *_Ad 2: _*
> The second document first published in July 2018 provides a very
> trivial to implement mechanism (without
> even changing BGP state machine) to automatically establish common LAN
> or p2p interfaces BGP sessions.
> Typical use case would be Clos DC fabrics, customer PE-CE LANs, TORs
> to Compute Nodes etc ...
> Comments, questions, feedback on both proposals all very welcome.
> Kind regards,
> Robert.
>
> _______________________________________________
> GROW mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/grow