BFD between two routers separated by switch

[shreeram-pardhy]

hi All,

I have a scenario where my two routers are separated by a switch. R1 === SW === R2. The links betwen all the devices are part of bundle interface. Routers are NCSxxx series and teh switch is C9300. Does this setup allow to support BOB and BLB? From what i have read, the routers need to be connected back to back to allow BLB and BOB working together. Any ideas how i can have BOB on my setup. I want to ensure that the BLB session does not go down when individual member links (which is actively carrying the BLB traffic ) of the bundle go down. At teh same time i want to ensure that if the member link ( link which is not carrying BLB traffic )goes down, the BFD is able to detect it.

[balaji.bandi]

check some use case here :

https://community.cisco.com/t5/service-providers-blogs/bfd-over-logical-bundle-blb-implementation-on-ncs5500-platforms/ba-p/3309345

[shreeram-pardhy]

hi balaji,

Thanks for your reply. I did see this link earlier, but it doesn't seem to mention anything about routers that are separated by a switch stack.  i have seen another cisco link which says that for BOB we need to connect the routers back to back.

Routing Configuration Guide for Cisco ASR 9000 Series Routers, IOS XR Release 6.9.x - Implementing BFD [Cisco ASR 9000 Series Aggregation Services Routers] - Cisco

To support BFD on bundle member links, be sure that the following requirements are met:

• The routers on either end of the bundle are connected back-to-back without a Layer 2 switch in between.

However, there is no such comment mentioned in the NCS540s implementation guide. 

i am not sure what Cisco guidelines are for BOB support between rotuers that are separated by Layer2 switch stack. 

[Giuseppe Larosa]

Hello @shreeram-pardhy ,

the document provided reports the differences between BFD over Bundle BoB and BFD over Logical Bundle BLB

>>

Differences between BFD over Bundle and BFD over Logical Bundle

BFD over Bundle (BoB) (RFC 7130) has a BFD session on each bundle member. The client is the bundle manager. If a BFD session goes down on a specific member link, the whole bundle interface goes down. That is, when the member link goes down, the number of available links falls below the required minimum. Hence the routing session is brought down.

BFD over Logical Bundle (BLB) (RFC 5880) treats a bundle interface with all its members as a single interface. BLB is a multipath (MP) single-hop session. If BLB is configured on a bundle there is only one single BFD session that is active. This implies that only one bundle member is being monitored by BFD at any given time. The client is one of the routing protocols. When BFD detects a failure, the client brings down the routing session.

For BoB the requirement for back to back connections comes from the fact that each BFD session on each link needs to stay on that link.

The presence of a L2 switch on the path between the two routers make possible depending on load balancing algorythms used on the L2 switch that the sessions may be mixed up .

Looking at RFC 7130

https://datatracker.ietf.org/doc/html/rfc7130#section-2.3

>>

If the Your Discriminator field is nonzero and a micro-BFD over a
      LAG session is found, the interface on which the micro-BFD control
      packet arrived MUST correspond to the interface associated with
      that session

>>

 This document defines the BFD control packets for each micro BFD
   session to be IP/UDP encapsulated as defined in [RFC5881], but with a
   new UDP destination port 6784.

>>

On Ethernet-based LAG member links, the destination Media Access
   Control (MAC) is the dedicated multicast MAC address
   01-00-5E-90-00-01 to be the immediate next hop.  This dedicated MAC
   address MUST be used for the initial BFD packets of a micro-BFD
   session when in the Down/AdminDown and Init states.  When a micro-BFD
   session is changing into the Up state, the first bfd.DetectMult
   packets in the Up state MUST be sent with the dedicated MAC.  For BFD
   packets in the Up state following the first bfd.DetectMult packets,
   the source MAC address from the received BFD packets for the session
   MAY be used instead of the dedicated MAC.

   All implementations MUST be able to send and receive BFD packets in
   Up state using the dedicated MAC address.

in short BoB sessions are UDP using a different UDP port then standard BFD sessions to avoid confusion.

The different sessions use different source MAC address and a different Discriminator Field.

Your switch is a L2 stack we can suppose you have two links from each router to each of the members of the L2 stack this may save you day because for the L2 switch it might be more convenient to switch frames between ports on the same member device rather then sending them over the stack .

You should test your setup and see what happens if the L2 switch does not swap the frames between ports on different member devices when a port is avaialable on the same member you are fine otherwise you can have issues.

Hope to help

Giuseppe

[shreeram-pardhy]

thanks for your inputs, Giuseppe. It's a very good point that you have mentioned above regarding hashing. I will check this in the lab setup.

[MHM Cisco World]

well I never try or see doc. but the BOB and BLB use micro BFD dont care if the two device connect back to back or via SW
so I think it work in your case. 

https://www.lastopinion.io/index.php/2024/03/19/bfd-over-lag-design/

MHM

[shreeram-pardhy · Accepted solution]

hey guys, just wanted to close this issue. I have tried the setup in the lab. The micro bfd sessions are coming up, but we start to see issues when any one of the link on either of the routers is shutdown. We see the bfd sessions (tied with the bundle interface) flap. Post that, the routers try to establish sessions again but its not an ideal one. Eg ( if R1's interface 1 is shutdown, then R2 will try to use both of its interfaces to establish a session with R1's interface 2 ). This is causing issues and just not working properly. In short, for micro bfd sessions to work properly, they must be connected directly without a switch stack in middle. Thank you all for your replies. I hope this post will help someone who may try similar setup.