In full/plain IP networks, additional iBGP sessions between non-adjacent routers can trigger traffic loops and must be avoided.
In MPLS networks it's not a problem (thanks to the transport labels).
By the way, IP or MPLS, adding iBGP session between adjacent routers is not a problem, it allows using the most direct path, actually ; something that is now available too over RR using BGP-ORR (BGP Optimal Route Reflection feature) – but it wasn't your question.
------------------------------
Olivier Benghozi
------------------------------
Original Message:
Sent: 02-07-2025 16:59
From: Anonymous
Subject: BGP RR Question
This message was posted by a user wishing to remain anonymous
Hello,
I believe I know the answer to this based on reading RFC 4456, but I wanted to confirm.
I'm running a BGP signaled MPLS network. We use multiple Route Reflectors for scalability - so all our routers peer with all the RR's. We're going to be slowly migrating what is currently a L2 ethernet ring by creating L2VPNs between adjacent nodes, and migrating the older L2 gear to pass over the L2VPNs. The issue is that until this ring is complete, a fiber cut/equipment failure in the new MPLS portion of the network would cause the sessions between the downstream routers and the RR's to fail, thus causing my point-to-point L2VPNs to fail as well.
To combat this, I'm trying to verify if it's allowed to still have direct peerings between non-RR's as long as all routers still peer with the RR's. And, if the connectivity to the RR's is lost as I mentioned, will this allow circuits between these directly connected neighbors to continue functioning? Do/Should I peer with all the routers on this new ring so there continues to be a full mesh should they lose access to the RR's - or can not do this at all?
Thank you in advance,