Hello Greg/ Nobo, Thanks for this thoughtful discussion, I would like to clarify the motivation of this draft:
1. There are scenarios where fault detection be initiated and managed from one end since the logic to failover (to a backup PW) is present on only on that end. The other end may participate in the PW establishment (signaling) or the PW may be provisioned but does not need to be actively involved in fault detection. Also, the action of failing over to a backup PW is made at one end. This motivated us to use SBFD instead of classical BFD. With classical BFD, though we could use echo in one direction, both ends need to actively participate in the BFD session (through Async and by allocating other BFD session resources). 2. However the circuit (PW) in use is a bi-directional construct as you point out and hence the text in Sec2.2.2. Thanks to Nobo for helping clarify this. Prasad From: Rtg-bfd [mailto:[email protected]] On Behalf Of Nobo Akiya Sent: Tuesday, March 24, 2015 7:36 AM To: 'Gregory Mirsky'; [email protected] Subject: RE: BFD Echo format Hi Greg, I absolutely agree on the S-BFD reverse path needing to take the corresponding PW. I peeked into draft-gp-pals-seamless-vccv, and I did find a text for this. 2.2.2. S-BFD Reflector Operation All point to point pseudowires are bidirectional, the S-BFD Reflector therefore reflects the S-BFD packet back to the Initiator using the VCCV channel of the reverse direction of the PW on which it was received. So, even though this key aspect was not described in the presentation, it looks like the authors did consider this. Thanks! -Nobo From: Gregory Mirsky [mailto:[email protected]] Sent: March-23-15 8:21 PM To: Nobo Akiya; [email protected]<mailto:[email protected]> Subject: RE: BFD Echo format Hi Nobo, thank you for your thoughtful and interesting, as always, comments. I've realized, after the session, that encapsulation of BFD Echo over LSP or PW MUST be different from what defined in RFC 5880. I think it must follow the same principle as set in RFC 5884 for Async BFD mode. And I agree that PW, unlike LSP (at least most LSP constructs), is a bi-directional construct. Thus, as I think of it, operators more interested in verifying data plane availability in both directions. If we propose two-way single-ended mechanism to check data path of the particular PW, then, I imagine, the return path MUST be in-band, not over IP path. I think that we need to note that in our discussion of S-BFD or BFD Echo over VCCV as a requirement. Then we can see whether extending BFD Echo to MPLS LSP and PW or using S-BFD bring better solution, particularly since we already have ICMP and LSP Ping over VCCV defined. Regards, Greg From: Nobo Akiya [mailto:[email protected]] Sent: Monday, March 23, 2015 6:19 PM To: Gregory Mirsky; [email protected]<mailto:[email protected]> Subject: RE: BFD Echo format Hi Greg, I believe this comment was in context of S-BFD for VCCV. >From what I understand, the scenario which S-BFD for VCCV is looked into is >the case where one end of the PW has very constrained resources. Using the >classical BFD will require such devices to maintain a session object per PW, >as opposed to S-BFD which only requires one reflector session. Your comment about why not send BFD echo packets over the PW? Sending a probe (BFD echo) with IP destination having the sender address, across multiple physical hops, always has the danger of the probe packet still coming back to the sender despite issues. Yes this is more likely on LSPs like LDP, RSVP (and less likely on PW) but it is possible that PW label gets exposed "somewhere" and that label value happens to match a different LSP, picking up traversal to a different node. Then there's the desire for return packets to come back on the reverse direction of the forward PW. Whether we go with S-BFD for VCCV or BFD echo, this (i.e., reflector bouncing back packets on the reverse PW) is a change that needs to be made/standardized. So to me, S-BFD for VCCV makes sense. Thanks! -Nobo From: Rtg-bfd [mailto:[email protected]]<mailto:[mailto:[email protected]]> On Behalf Of Gregory Mirsky Sent: March-23-15 1:54 PM To: [email protected]<mailto:[email protected]> Subject: BFD Echo format Dear All, I stand corrected, BFD Echo format is not defined in the RFC 5880 but there requirement suggesting that BFD Discriminator MAY be used to demux Echo replies: The payload of a BFD Echo packet is a local matter, since only the sending system ever processes the content. The only requirement is that sufficient information is included to demultiplex the received packet to the correct BFD session after it is looped back to the sender. The contents are otherwise outside the scope of this specification. Regards, Greg
