[Final 2020 Problem 2.1.c] BGP injection of routes learnt via E-BGP

[Final 2020 Problem 2.1.c] BGP injection of routes learnt via E-BGP

by Tom Louis Demont -
Number of replies: 6
Hello,
For this part of the problem, I'm having troubles understanding why R2 does not receive announcements from B1 and B2 for the routes they learnt from E-BGP? I understand B1 and B2 won't select those routes as their best route but depending on the timing, couldn't we have that B3 and B4 give their updates to B1 and B2 a bit later and therefore, B3 and B4 would already have announced to R2 their Best Route to C (that in that case goes through A)? I'm not sure I understand the mechanism that would prevent this kind of behaviour to happen.
More generally, is there any specific propagation order we should assume for the I-BGP routes announcements?

Thanks for your help.
In reply to Tom Louis Demont

Re: [Final 2020 Problem 2.1.c] BGP injection of routes learnt via E-BGP

by Stéphan Plassart -

Dear Tom,

Here there is no timing issue, because in the question we assume that all routing protocol have converged. Indeed, in the statement of question1, it is specified that:  "At time t1, BGP and OSPF have converged in all ASs.".

Hope this answer your interrogation,

Best regards,

Stéphan.

In reply to Stéphan Plassart

Re: [Final 2020 Problem 2.1.c] BGP injection of routes learnt via E-BGP

by Tom Louis Demont -
Hello,
I understand that routing protocols are said to have converged, but what would prevent B1 and B2 from having sent the routes they learnt from E-BGP :
dest = 4001:1::/32, next-hop = a2e, AS-Path = A-C

or

dest = 4001:1::/32, next-hop = a1e, AS-Path = A-C

although these are not the preferred routes, these are still routes received by B1 and B2. If during the BGP exchanges, B1 and B2 would have had time to send these announcements (that could have been the best routes for B1/B2 before B3 and B4 made internal announcements), wouldn't R2 have in its RIB-adj-in table those announcements?

Thanks again for your time

In reply to Stéphan Plassart

Re: [Final 2020 Problem 2.1.c] BGP injection of routes learnt via E-BGP

by Devrim Celik -
Hello,

I think I had the same question while doing the task. When asking to list all the received packets, the way I understand it, this is a question during the convergences process. Why do I think that? Because BGP only sends additions and withdrawals, i.e., if the question is asked after convergence, since there is no modification of the links at t1, there will be no more communication and no routes will be received.
(If this is correct until this point...) This means, that we are asking for routes during the convergence process; and if we do that, using the same argument that Tom made, routes that may turn out to be sub-optimal in the long run, can be send and be received during this process.
In reply to Tom Louis Demont

Re: [Final 2020 Problem 2.1.c] BGP injection of routes learnt via E-BGP

by Plouton Grammatikos -
Dear Tom,

It is true that during the convergence process, B1 might advertise to R2 a route as a best one, that later turns out not to be the best one. If this happens, then B1 will send a withdraw message to R2 saying that the route advertised before is no longer best, so R2 should delete it. What the question asks is what are the BGP routes in R2 at time t1, when BGP has converged, not all the routes that might have been received by R2 during the convergence process.

I hope this helps clarify the question.

Best,
Plouton
In reply to Plouton Grammatikos

Re: [Final 2020 Problem 2.1.c] BGP injection of routes learnt via E-BGP

by Devrim Celik -
Hello Plouton,

I think the confusion we had comes from the fact that they asked for all the "received" ones; thank you for the clarification.

Best
Devrim
In reply to Plouton Grammatikos

Re: [Final 2020 Problem 2.1.c] BGP injection of routes learnt via E-BGP

by Tom Louis Demont -
Hello,
Thanks a lot, this helps a lot to clarify, I indeed haven't thought that the routes would be withdrawn, that's the mechanism I was missing