BFD between FRR and Cisco NX-OS does not work

[micush]

Description

At the title says, BFD between FRR and Cisco NX-OS does not work. It works between various other OSes, just not NX-OS.

Version

10.2.1

How to reproduce

NX-OS:
router bgp 65001
neighbor 192.168.1.2
remote-as 65002
bfd

FRR:
router bgp 65002
neighbor 192.168.1.1 remote-as 65001
neighbor 192.168.1.1 bfd

Expected behavior

I expect the hosts to become neighbors

Actual behavior

The hosts do not become neighbors

Additional context

No response

Checklist

• I have searched the open issues for this bug.
• I have not included sensitive information in this report.

[donaldsharp]

This is really nothing we can do here with this at all. We don't have any nxos devices. You are going to have to provide some useful data. Can you provide debugs from both sides? Can you provide a tcpdump of the failed connections?

[ton31337]

Could you try upgrading to 10.2.2? There are a couple important fixes regarding BFD. Also, could you share the logs and/or a pcap file?

[micush]

Here's a pcap from 10.2.1. I'll try again with 10.2.2.

nobfd.pcapng.gz

[micush]

I upgraded to 10.3 and am still seeing the same behavior.

[ton31337]

In your pcap... Which IP is FRR, which is NX-OS?

[micush]

10.0.0.8 is FRR

[lukedirtwalker]

We used containerlab and could reproduce the same issue.

What is interesting if on the FRR side you only configure bfd and have no "neighbor x.x.x.x bfd" entry in the bgp section the BFD seems to establish without problem.

We tried various different configurations and apart from no configuring bfd in the BGP section (as mentioned above) nothing seems to work consistently in FRR 10.2.2.
With FRR 9.1.3 everything works without issues.

A detailed report of what we tested is listed here: https://docs.google.com/document/d/1HiH70q2IEx-eH4aJJ1vvGlTadwqBk0eEyfhZLQVjLuw/edit?tab=t.0

I'm happy to help to debug this issue more. If you need any information to better pin-point the problem please let me know.

[micush]

This is great, thanks. I've also tested against ArubaOS-CX with the same results. No BFD neighbors there either.

[micush]

I confirm the findings of @lukedirtwalker . It works as expected when unconfigured in BGP and only configured in BFD:

router bgp 4216800522
 bgp bestpath as-path multipath-relax
 neighbor 192.168.5.252 remote-as 4216825566
 address-family ipv4 unicast
  redistribute connected
  neighbor 192.168.5.252 addpath-tx-all-paths
  neighbor 192.168.5.252 soft-reconfiguration inbound
  maximum-paths 4
 exit-address-family
!
bfd
 peer 192.168.5.252 local-address 192.168.5.12
 exit
!
!
# show bfd peers brief
Session count: 1
SessionId  LocalAddress                             PeerAddress                             Status         
=========  ============                             ===========                             ======         
2216826743 192.168.5.12                             192.168.5.252                           up             
!

[ton31337]

@lukedirtwalker as I understand 9.1.3 is the last "working" version?

[lukedirtwalker]

@lukedirtwalker as I understand 9.1.3 is the last "working" version?

We specifically tested 10.2.2 and 10.2.1 and confirmed them to be not working.
And respectively 9.1.3 to be working.

If you want I can easily spin up a test for a different 10.x version. Which one(s) should I try?