Hi Yunan, > Regarding your suggestion of adding a "ECMP" path type, well, the currently > defined "0x0004 -- Primary Path" path type should do the work. In fact, the > so-called "Primary Path" in this draft refers to all the ECMP paths > (including the "Best path"). Of course, we can further work on the naming.
Just to clarify. There can be multiple primary paths and one of it will be the best path as well. Correct? If yes perfect. Thank you. > May I ask the use case for your proposal of adding "the reason why it is > considered ECMP" reason string? When troubleshooting BGP, one of the most use cases are why a path is not installed for forwarding. One reason can be because of various configuration knobs such as: Configuration Guide - IP Unicast Routing - BGP Configuration - Configuring BGP Load Balancing https://support.huawei.com/enterprise/en/doc/EDOC1100004198/19f9347b/configuring-bgp-load-balancing - maximum load-balancing (maximum-path) - load-balancing as-path-relax At the end we will search within BMP local-RIB collected metrics the same way as you would do on CLI described in this document. The big difference is that we will do it across all the routing contexts and across the network. Thus, much more efficiently. Configuration Guide - IP Unicast Routing - BGP Configuration - Verifying the BGP Route Selection and Load Balancing Configuration https://support.huawei.com/enterprise/en/doc/EDOC1100004198/6efe04ed/verifying-the-bgp-route-selection-and-load-balancing-configuration I always have the closed loop operation in mind as well. What BMP metrics do I need so that I can trigger a configuration change to achieve the desired intend. The intend is that I want a certain selection of paths be ECMP. Two possible use cases can happen: - To many are installed. Some of them are undesired. - To less are installed. Some should but aren't. If they are not installed, we should see them in "Non-installed path" or perhaps "Backup path" or "Best external path". > Do you have any specific reason string in mind for the ECMP paths? If they are installed then it would be great to understand in the ECMP decision process of the router within the routing context which was the last decision step. For instance: - AS-PATH relax - eBGP multipath - iBGP multipath - eiBGP multipath draft-lapukhov-bgp-ecmp-considerations-02 shows what possibilities we have. If we see a route marked as ECMP with "eBGP multipath" but was not intend to be part of ECMP, we could remove the configuration line "maximum load-balancing ebgp" and verify if desired route is now marked as "Non-installed path" or perhaps "Backup path" or "Best external path". I hope this makes sense. Feedback very welcome. Kind regards Thomas Graf ____________________________________________________________________________ Network Engineer Datacenter Functions Telefon +41-58-223 84 01 Mobile +41-79-728 80 12 thomas.g...@swisscom.com ____________________________________________________________________________ Swisscom (Schweiz) AG IT, Network & Infrastructure Datacenter Functions Binzring 17 8045 Zürich www.swisscom.com Postadresse: Binzring 17 8045 Zürich -----Original Message----- From: Guyunan (Yunan Gu, IP Technology Research Dept. NW) <guyu...@huawei.com> Sent: Monday, July 15, 2019 11:52 AM To: Graf Thomas, INI-ONE-WSN-DCF <thomas.g...@swisscom.com>; juancamilo.card...@imdea.org; grow@ietf.org; pa...@ntt.net Cc: draft-cppy-grow-bmp-path-marking-...@ietf.org Subject: RE: [GROW] Path marking using BMP - TLVs Hi Thomas, Thanks a lot for your comments on the draft, and your elaboration on the usage of Path Marking TLV! Regarding your suggestion of adding a "ECMP" path type, well, the currently defined "0x0004 -- Primary Path" path type should do the work. In fact, the so-called "Primary Path" in this draft refers to all the ECMP paths (including the "Best path"). Of course, we can further work on the naming. Thanks for sharing the ECMP reference: raft-lapukhov-bgp-ecmp-considerations-02. May I ask the use case for your proposal of adding "the reason why it is considered ECMP" reason string? Well, it's easy for me to understand that users may wonder why a path is "non-best". Do you have any specific reason string in mind for the ECMP paths? BR, Yunan -----Original Message----- From: thomas.g...@swisscom.com [mailto:thomas.g...@swisscom.com] Sent: Friday, July 12, 2019 8:44 PM To: juancamilo.card...@imdea.org; grow@ietf.org; pa...@ntt.net; Guyunan (Yunan Gu, IP Technology Research Dept. NW) <guyu...@huawei.com> Cc: draft-cppy-grow-bmp-path-marking-...@ietf.org Subject: RE: [GROW] Path marking using BMP - TLVs Hi Camilo, Paulo and Yunan, Thank you very much for this exciting and very useful draft. This will make draft-ietf-grow-bmp-local-rib even more useful. On top of having access to all (not only to the best) BGP paths in BGP local RIB, thanks to this draft, we will finally understand how these BGP paths are installed in RIB/FIB. We will be able to get an network wide overview on which routers which paths have redundancy and which not. And that with ONE single query in big data. One remark I like to add to complete all the possible path types. In section 2.1 Path Type, could you add ECMP (Equal-Cost Multipath) as path type and under Section 2.2 Reason String, the reason why it is considered ECMP. For BGP Equal-Cost Multipath reasons, please refer to this current draft: https://www.ietf.org/id/draft-lapukhov-bgp-ecmp-considerations-02.txt Kind regards Thomas Graf ____________________________________________________________________________ Network Engineer Datacenter Functions Telefon +41-58-223 84 01 Mobile +41-79-728 80 12 thomas.g...@swisscom.com ____________________________________________________________________________ Swisscom (Schweiz) AG IT, Network & Infrastructure Datacenter Functions Binzring 17 8045 Zürich www.swisscom.com Postadresse: Binzring 17 8045 Zürich -----Original Message----- From: GROW <grow-boun...@ietf.org> On Behalf Of Camilo Cardona Sent: Saturday, July 6, 2019 5:04 AM To: grow@ietf.org grow@ietf.org <grow@ietf.org> Cc: draft-cppy-grow-bmp-path-marking-...@ietf.org Subject: [GROW] Path marking using BMP - TLVs Hello GROW, We just submitted draft https://www.ietf.org/id/draft-cppy-grow-bmp-path-marking-tlv-00.txt. The idea of the draft is to signal the state of the path in the FIB using the mechanism described in draft-lucente-bmp-tlv-00 (https://www.ietf.org/id/draft-lucente-bmp-tlv-00.txt), which was introduced this week. Feedback is, as always, welcome. If possible, we would like to have a couple of minutes to present it in Montreal (probably better if done next to the presentation of draft-lucente-bmp-tlv-00). A good part of this document was inspired by other draft, https://tools.ietf.org/html/draft-bgp-path-marking-00, that we proposed some years ago. In that draft, similar information was signaled using communities. Back then, there were some concerns of this data potentially messing with the BGP decision process, something that should not be a problem when using BMP. Thanks, Camilo Cardona _______________________________________________ GROW mailing list GROW@ietf.org https://www.ietf.org/mailman/listinfo/grow
smime.p7s
Description: S/MIME Cryptographic Signature
_______________________________________________ GROW mailing list GROW@ietf.org https://www.ietf.org/mailman/listinfo/grow