Hi, Gunter: I don’t recommend to design one " brand-new control plane signal " to signal the prefix reachability.
What I argue with you until now, is the following statements: "when a UPA stops being advertised, that unreachability signal goes away, and at that point, the summary route takes over and signals reachability." >From the UPA document, we can know there at least two reasons can lead the >"UPA stops being advertised": Case 1: "UPA advertisements SHOULD therefore be withdrawn after some amount of time, that would provide sufficient time for UPA to be flooded network-wide and acted upon by receiving nodes, but limits the presence of UPA in the network." Case 2: " ABR or ASBR MUST withdraw the previously advertised UPA when the reason for which the UPA was generated was lost " Then, if the reason that lead " UPA stops being advertised " is Case 1, then your logic that " at that point, the summary route takes over and signals reachability." is wrong. When the BGP PIC FRR switch back at this moment, the traffic will be lost. Best Regards Aijun Wang China Telecom -----邮件原件----- 发件人: [email protected] [mailto:[email protected]] 代表 Gunter van de Velde (Nokia) 发送时间: 2025年5月28日 15:04 收件人: Aijun Wang <[email protected]> 抄送: [email protected]; 'Peter Psenak' <[email protected]> 主题: [Lsr] Re: 答复: I-D Action: draft-ietf-lsr-igp-ureach-prefix-announce-06.txt [speaking as work group participant] Hi Aijun, Let’s wrap up this thread. I feel we’ve explored this topic quite thoroughly from different angles, and I’d prefer not to keep circling back and repeating the same points. (this is my last email on this thread) Regarding your comment: " 1) The data-plane OAM tools(BFD, ICMP, STAMP, and TWAMP) should be used to test and verify the prefix reachability. Along the evolution of PUAM/UPA drafts, there are lots of discussions among the efficiency of these OAM tools, versus the PUAM/UPA mechanism. I think you may miss these previous discussions. Here I want just to say, if the WG agree with you, the PUAM/UPA draft shouldn't be existed. " Let me try to explain my point again, one last time, just in a different way for clarity. A router forwards packets based on the unicast forwarding table, which is typically populated by routing protocols like OSPF or IS-IS. Once a route is selected, the router uses that information to forward packets, plain and simple. But here’s the thing: the router doesn’t really “know” if the packet will make it all the way to its final destination. It just follows the best path it has available, based on its local view. What happens downstream, packet drops, blackholing, filtering, or misrouting. can’t be predicted by the control plane alone. That’s just the reality of distributed routing. At the end of the day, if we really want to be sure something is reachable and working as expected, the best way to confirm that is to actually test it. Tools like BFD, ICMP, STAMP, and TWAMP are exactly what we use to validate that, since they measure actual behavior in the data plane. >From your comments, it sounds like you're proposing a new control plane signal >that says, “Hey, trust me, I’m really reachable.” But I’m not convinced such a >signal would be any more reliable than the routes we already learn through >established routing protocols like OSPF or IS-IS. That kind of new control >plane signal could still be based on outdated, incorrect, or incomplete >information, and wouldn’t really help when it comes to truly assessing data >plane reachability. Coming back to UPA. When using UPA the procedure as described in [1], [2], and [3] is predictable and behaves in an operationally efficient way. That way, in most cases, we don’t even need to rely on OAM data plane testing tools. From my point of view, adding a new control plane message to somehow validate reachability on top of an already installed unicast route doesn’t really add much value. Instead, it risks introducing more complexity without solving a real problem. Regarding your comment: " 2) In BGP PIC FRR scenario, when the router stops receiving the UPA signal, and "switching back to the primary egress router" is wrong. As stated in your reply, stop receiving the UPA signal " ...doesn’t imply anything about whether the prefix is reachable ", then, when the prefix is still unreachable, switch back to the primary egress router will result in the traffic black hole. " GV> If the UPA disappears then the assumption is that the prefix is indeed reachable, assuming the prefix would be a member prefix covered by the summary route. See [1],[2] and [3] for the procedure. This is basic network behavior when using aggregate routes on ABRs. If this is not the behavior desired, then an alternative is to not summarize prefixes into aggregate prefixes (but this has other operational implications). " In summary, current UPA proposal try to signal the unreachability on demand, but lack of the signaling mechanism to revoke such announcement on demand. " GV> What I’m saying is pretty straightforward: when a UPA stops being advertised, that unreachability signal goes away, and at that point, the summary route takes over and signals reachability. Summary routes have been used in networks for years and has worked reliably. That’s why I’m having a hard time understanding how introducing a brand-new control plane message to signal destination reachability would actually help. It doesn’t really add anything new when it comes to confirming data plane reachability, and we already have mechanisms that work well in practice. Note that this was my last email on this subject as I think we now discussed this enough. Brgds, G/ [speaking as work group participant] [1] https://mailarchive.ietf.org/arch/msg/lsr/EFb71fc-hc-_oMCnksOZfSPt0-o/ {2] https://mailarchive.ietf.org/arch/msg/lsr/aAyfU1ziWTuTltmZoCGO1vht7Qk/ {3] https://mailarchive.ietf.org/arch/msg/lsr/hg_v2ZAOkz5bhxE4979viIowkIU/ -----Original Message----- From: Aijun Wang <[email protected]> Sent: Wednesday, May 28, 2025 3:06 AM To: Gunter van de Velde (Nokia) <[email protected]> Cc: [email protected]; 'Peter Psenak' <[email protected]> Subject: 答复: [Lsr] Re: 答复: I-D Action: draft-ietf-lsr-igp-ureach-prefix-announce-06.txt CAUTION: This is an external email. Please be very careful when clicking links or opening attachments. See the URL nok.it/ext for additional information. Hi, Gunter: There are two misunderstanding in your arguments: 1) The data-plane OAM tools(BFD, ICMP, STAMP, and TWAMP) should be used to test and verify the prefix reachability. Along the evolution of PUAM/UPA drafts, there are lots of discussions among the efficiency of these OAM tools, versus the PUAM/UPA mechanism. I think you may miss these previous discussions. Here I want just to say, if the WG agree with you, the PUAM/UPA draft shouldn't be existed. 2) In BGP PIC FRR scenario, when the router stops receiving the UPA signal, and "switching back to the primary egress router" is wrong. As stated in your reply, stop receiving the UPA signal " ...doesn’t imply anything about whether the prefix is reachable ", then, when the prefix is still unreachable, switch back to the primary egress router will result in the traffic black hole. In summary, current UPA proposal try to signal the unreachability on demand, but lack of the signaling mechanism to revoke such announcement on demand. Best Regards Aijun Wang China Telecom -----邮件原件----- 发件人: Gunter van de Velde (Nokia) [mailto:[email protected]] 发送时间: 2025年5月27日 17:39 收件人: Aijun Wang <[email protected]> 抄送: [email protected]; 'Peter Psenak' <[email protected]> 主题: RE: [Lsr] Re: 答复: I-D Action: draft-ietf-lsr-igp-ureach-prefix-announce-06.txt [speaking as work group participant] A router makes its forwarding decisions based on the routes it has in the unicast forwarding table, which are populated by the control plane. When a packet comes in, the router looks up the destination, and if there’s a match, it uses the forwarding info tied to the best route it has at that moment to send the packet on its way. It is worth pointing out that the router doesn’t actually know whether the packet will make it all the way to the final destination. It simply follows the best available path based on local routing information. Along the way, anything from packet filtering, DDoS mitigation, or buffer drops could affect delivery, that’s just the nature of network behavior. If there are concerns about whether a specific destination is actually reachable, that’s where data-plane OAM tools come in. Things like BFD, ICMP, STAMP, and TWAMP are built specifically to test and verify reachability beyond what the control plane can determine. To clarify something mentioned earlier when I said “when the UPA is no longer advertised and only the summary prefix remains, the system quickly and gracefully reverts back to the original state without any issues,” I meant exactly that. It doesn’t imply anything about whether the prefix is reachable, it just means the system sees the UPA is gone, and so it reverts to whatever route is considered best in the routing table. In the case of BGP PIC FRR, that usually means switching back to the primary egress router, depending on how the feature is configured. Whether it actually switches back or sticks with the alternate path depends on the specific BGP PIC settings and current network state. G/ (speaking as WG participant) -----Original Message----- From: Aijun Wang <[email protected]> Sent: Tuesday, May 27, 2025 10:47 AM To: Gunter van de Velde (Nokia) <[email protected]> Cc: [email protected]; 'Peter Psenak' <[email protected]> Subject: 答复: [Lsr] Re: 答复: I-D Action: draft-ietf-lsr-igp-ureach-prefix-announce-06.txt CAUTION: This is an external email. Please be very careful when clicking links or opening attachments. See the URL nok.it/ext for additional information. Hi, Gunter: Here is the tricky issue: "when the UPA is no longer advertised and only the summary prefix remains, the system quickly and gracefully reverts back to the original state without any issues." Then, you assumes that: when no UPA is advertised, the specific prefix is reachable again? Best Regards Aijun Wang China Telecom -----邮件原件----- 发件人: Gunter van de Velde (Nokia) [mailto:[email protected]] 发送时间: 2025年5月27日 16:17 收件人: Aijun Wang <[email protected]> 抄送: [email protected]; Peter Psenak <[email protected]> 主题: RE: [Lsr] Re: 答复: I-D Action: draft-ietf-lsr-igp-ureach-prefix-announce-06.txt [speaking as work group participant] Hi Aijun, I just wanted to share a couple of thoughts in response to your message. First, I feel your note is revisiting the topic around the PUA draft [2] in a way that runs counter to the earlier formal warning from Yingzhen [1]. Reopening that discussion at this point doesn’t help strengthen the UPA procedures and seems to be circling back on a topic that the working group has already moved past. Second, based on the procedures outlined in the UPA draft [3], my development team has successfully implemented UPA and is using it as foundational part of our BGP PIC FRR solution. We’re quite familiar with how link-state protocols operate, and we’ve found the documented behavior in [3] to be sufficient for developing an interoperable and reliable implementation. In our solution, the UPA is used to trigger BGP PIC fast reroute exactly as Peter described [4]. When a UPA is received, the BGP PIC FRR process kicks in, and when the UPA is no longer advertised and only the summary prefix remains, the system quickly and gracefully reverts back to the original state without any issues. This solution is now used by network operators who understand the procedures of link-state routing protocols, UPA and BGP PIC FRR well. [1] https://mailarchive.ietf.org/arch/msg/lsr/S7Mqk2JljYgG8kjEPmH6KlLtbfw/ [2] https://datatracker.ietf.org/doc/draft-wang-lsr-prefix-unreachable-annoucement/ [3] https://datatracker.ietf.org/doc/draft-ietf-lsr-igp-ureach-prefix-announce/ [4] https://mailarchive.ietf.org/arch/msg/lsr/EFb71fc-hc-_oMCnksOZfSPt0-o/ Hope this helps clarify our experience and perspective. Gunter Van de Velde (speaking as WG participant) From: Aijun Wang <[email protected]> Sent: Tuesday, May 27, 2025 1:05 AM To: Peter Psenak <[email protected]> Cc: [email protected] Subject: [Lsr] Re: 答复: I-D Action: draft-ietf-lsr-igp-ureach-prefix-announce-06.txt CAUTION: This is an external email. Please be very careful when clicking links or opening attachments. See the URL nok.it/ext for additional information. Hi, Peter: Please refrain your comments. As the initiators of such mechanism, we describe the whole process 3 years earlier than the first version of your draft. Respect the creator of any idea is the basic behavior of IETF community. Back to your example: You forgot the important part of the puzzle—-the application that the UPA triggered. How and when the application being triggered back to its original state? Image one scenario, some nodes advertised accidentally the wrong UPA signals within the network, how can you revoke it quickly and eliminate the wrong triggered actions? Aijun Wang China Telecom On May 26, 2025, at 17:52, Peter Psenak <mailto:[email protected]> wrote: On 26/05/2025 10:26, Aijun Wang wrote: No, summary can’t achieve the same aim of the “Explicit Withdrawn Signal”, for example, switch back to the application’s original state. you don't understand the basic operation of the protocol. 1. prefix p1/32 is summarized with p2/16. P1 is reachable via summary 2. router that generated p1 went down 3. UPA for p1/32 was generated 4. router that generated p1 came back 5. UPA was removed and we are back to state (1) Peter 发件人: Peter Psenak [mailto:[email protected]] 发送时间: 2025年5月26日 15:41 收件人: Aijun Wang mailto:[email protected] 抄送: mailto:[email protected] 主题: Re: 答复: [Lsr] Re: 答复: I-D Action: draft-ietf-lsr-igp-ureach-prefix-announce-06.txt On 26/05/2025 03:29, Aijun Wang wrote: Then one new deficiency for the mechanism is emerging: The lack of the Explicit Withdrawn Signal(EWS) when the prefix is reachable again. Please note, stop sending the UPA message doesn’t mean the prefix is reachable again. If there is no EWS, then the network can’t back to its original state before the UPA signaling when the reachable of prefix recover. there is still a summary that covers the prefix reachability. Peter Best Regards Aijun Wang China Telecom 发件人: Peter Psenak [mailto:[email protected]] 发送时间: 2025年5月23日 19:11 收件人: Aijun Wang mailto:[email protected] 抄送: mailto:[email protected] 主题: Re: [Lsr] Re: 答复: I-D Action: draft-ietf-lsr-igp-ureach-prefix-announce-06.txt On 23/05/2025 12:48, Aijun Wang wrote: Then nothing needs to be standardized when the prefix becomes reachable again. 1) In some critical scenarios, when the ABR sends one UPA message out and the prefix becomes reachable immediately, what the ABR can do is to stop advertising UPA. and that is exactly what the text says. Peter The sent UPA message will eventually trigger the action on the receiver, even the prefix is reachable immediately. 2) In normal situations, the ABR sends the UPA message for some time and stop sending it further. At this time, when the prefix becomes reachable, nothing needs to be done at ABR. The receiver will also act on the UPA signaling. It’s irrelevant then whether the prefix is reachable or not after the UPA signaling is sent out. Aijun Wang China Telecom On May 23, 2025, at 17:18, Peter Psenak mailto:[email protected] wrote: On 23/05/2025 10:10, Aijun Wang wrote: Then, what’s the differences between the two statements: first is for case when the prefix reachability is not regained after UPA was generated. Second is when the prefix reachability was regained before the UPA was withdrawn. It basically says UPA must be withdrawn at the time the prefix becomes reachable. “UPA advertisements SHOULD therefore be withdrawn after some amount of time, that would provides sufficient time for UPA to be flooded network-wide and acted upon by receiving nodes, but limits the presence of UPA in the network.” And: “ABR or ASBR MUST withdraw the previously advertised UPA when the reason for which the UPA was generated was lost - e.g. prefix reachability was restored or its metric has changed such that it does not represent the protocol specific maximum prefix metric.” Here, does “withdraw” just mean to “stop advertisement”? yes. Peter If no, what’s the mechanism of second “withdraw”? Best Regards Aijun Wang China Telecom 发件人: mailto:[email protected] [mailto:[email protected]] 代表 Peter Psenak 发送时间: 2025年5月23日 14:55 收件人: Aijun Wang mailto:[email protected]; mailto:[email protected] 主题: [Lsr] Re: 答复: I-D Action: draft-ietf-lsr-igp-ureach-prefix-announce-06.txt On 23/05/2025 03:32, Aijun Wang wrote: Hi, All: I must point out that the updated draft doesn't address previous issues that described in [1]. Especially, the activation of flawed LSInfinity feature(there is detail analysis for this flawed feature that is defined in OSPF 2328). And, some updated contents will deteriorate the traffic pattern within the network. For example, It says: “ABR or ASBR MUST withdraw the previously advertised UPA when the reason for which the UPA was generated was lost”. The above requirement will advertise the specific prefixes within the network, which will weaken the original summary effect, and attract the traffic via one or some of ABRs. no, above is not true, the new text does not say to advertise reachablity for a summarized prefix, it only talks about removing the previously advertised UPA. Please read carefully before commenting. Peter [1]: Reasons of abandoning UPA: https://datatracker.ietf.org/doc/draft-wang-lsr-reasons-of-abandon-upa-proposal/ Best Regards Aijun Wang China Telecom -----邮件原件----- 发件人: mailto:[email protected] [mailto:[email protected]] 代表 mailto:[email protected] 发送时间: 2025年5月22日 21:20 收件人: mailto:[email protected] 抄送: mailto:[email protected] 主题: [Lsr] I-D Action: draft-ietf-lsr-igp-ureach-prefix-announce-06.txt Internet-Draft draft-ietf-lsr-igp-ureach-prefix-announce-06.txt is now available. It is a work item of the Link State Routing (LSR) WG of the IETF. Title: IGP Unreachable Prefix Announcement Authors: Peter Psenak Clarence Filsfils Daniel Voyer Shraddha Hegde Gyan Mishra Name: draft-ietf-lsr-igp-ureach-prefix-announce-06.txt Pages: 14 Dates: 2025-05-22 Abstract: In the presence of summarization, there is a need to signal loss of reachability to an individual prefix covered by the summary. This enables fast convergence by steering traffic away from the node which owns the prefix and is no longer reachable. This document describes how to use the existing protocol mechanisms in IS-IS and OSPF, together with the two new flags, to advertise such prefix reachability loss. The IETF datatracker status page for this Internet-Draft is: https://datatracker.ietf.org/doc/draft-ietf-lsr-igp-ureach-prefix-announce/ There is also an HTMLized version available at: https://datatracker.ietf.org/doc/html/draft-ietf-lsr-igp-ureach-prefix-announce-06 A diff from the previous version is available at: https://author-tools.ietf.org/iddiff?url2=draft-ietf-lsr-igp-ureach-prefix-announce-06 Internet-Drafts are also available by rsync at: rsync.ietf.org::internet-drafts _______________________________________________ Lsr mailing list -- mailto:[email protected] To unsubscribe send an email to mailto:[email protected] _______________________________________________ Lsr mailing list -- mailto:[email protected] To unsubscribe send an email to mailto:[email protected] _______________________________________________ Lsr mailing list -- mailto:[email protected] To unsubscribe send an email to mailto:[email protected] _______________________________________________ Lsr mailing list -- [email protected] To unsubscribe send an email to [email protected] _______________________________________________ Lsr mailing list -- [email protected] To unsubscribe send an email to [email protected]
