Spine&leaf BGP unnumbered and VRF, subinterfaces on interlinks

  • 24 October 2017
  • 4 replies

So, I found this "leaf spine vrf bgp" topic: https://getsatisfaction.cumulusnetworks.com/cumulus/topics/leaf-spine-vrf-bgp.
I wonder if it's still the only and appropriate way to implement this functionality, which is to let spines and leaves establish BGP neighborship. My fresh experience with Cumulus Linux teaches that something might have changed since the topic mentioned below (it's 7 months old).

4 replies

Userlevel 3
Hi Jakub, The design from that post is definitely still valid and running in customer networks. Can you elaborate on what is not working as expected?
Why do You anticipate something not working? 😉 I just started playing with vrf functionality and it looks promising. After initial fail, the post helped me getting on the right track.
Knowing the pace of CL devlopment and all the new improvements, I just wonder if anything new arrived to Cumulus Linux that changed the situation.
I marked my post as "question", not a "problem", I hope.
Userlevel 3
LOL...Occupational hazard. Being a support engineer, we immediately want to fix things!
The VRF work we have added to the linux kernel is awesome; bringing linux networking to a new level. We are currently working on MPLS features (I think these are targeted for 4.0), which hopefully includes Layer3-VPN (I am not sure what is on the roadmap currently). This would definitely be an improvement on the current VRF-lite method of multi-VRF leafs.
O! And just before Your reply I hit the [CS#5404] on one of my spines. Just after I verified routes from leaves are appearing in vrf routing table on spines. What a coincidence, You were supporting enginner on that request! Now I'm waiting for spine2 to fail, sad moment!