Hi Jorge, Please refer to my comments inline marked w/ "Ali>"
On 3/21/18, 1:59 AM, "Rabadan, Jorge (Nokia - US/Mountain View)" <[email protected]> wrote: Ali and authors, As discussed during the BESS session, these are the points that I think should be addressed in draft-ietf-bess-evpn-igmp-mld-proxy before WG LC: -------------------------------------------------------------------------------- 1) Fast Leave text addition There are quite a few igmp-snooping implementations in the market that support a “Fast Leave” mechanism. EVPN should incorporate/document this too, since it is a pretty common use-case. Implementations allow the use of "Fast Leave" when the IGMP host is directly connected to the PE/NVE or the directly connected CE does igmp-proxy (and only in those cases). Fast Leave is a local administrative option on each AC, that, if enabled, allows the removal of a (x,G) state immediately after the reception of an IGMP Leave message for the (x,G). Ali> But the option of "fast-leave" requires for the PE to do host tracking and in case of IGMPv2, if there are more than one hosts is sitting behind an AC, it is difficult to do host-tracking because of the report suppression in IGMPv2 !! So, if used, it needs to be used with caution for only a single host for IGMPv2. I can add this "fast-tracking" as an option (MAY) but with the caveats that it has !! In the email below, I was suggesting that in some cases the IGMP Leave synch route can be avoided; however Mankamana made me see that, in the Fast Leave procedure, the PE receiving the IGMP Leave on the ES' AC, should always send an IGMP Leave sync route with an indication that the (x,G) state must be removed immediately. Mankamana suggested MRT=0 (Max Response Time=0) in the route could give that indication to the other PEs in the ES. Ali> Yes, fast-leave (if used) still need to be synchronized among multi-homing PEs ( Authors, can you please add text about Fast Leave? Ali> Since majority of existing implementation support "fast-leave", we can add it as an option (MAY) with the caveats that I described above. -------------------------------------------------------------------------------- 2) Conflicting text about advertising SMET route when there are local sources "3.2 PE with mixed of attached hosts/VMs and multicast source The main difference in here is that when PE2 receives IGMPv3 Join from H7 for (S2,G2), it does not advertises it in BGP because PE2 knows that S2 is attached to its local AC." [JORGE] the above is contradicting this previous statement: "When the first hop PE receives an IGMPv3 Join for (S,G) on a given BD, it advertises the corresponding EVPN Selective Multicast Ethernet Tag (SMET) route regardless of whether the source (S) is attached to itself or not in order to facilitate the source move in the future." [JORGE] I tend to agree with the latter statement. It simplifies the procedure. Ali> Since EVPN inherently supports workload mobility, the latter should be the default mode of operation. I guess, we can have an option (MAY) for the former one. -------------------------------------------------------------------------------- 3) Confusing text in section 7.1.1 about local-bias: "The Originator Router Address is the IP address of Router Originating the prefix. It should be noted that using the "Originating Router's IP address" field is needed for local-bias procedures and may be needed for building inter-AS multicast underlay tunnels where BGP next hop can get over written." While I agree with the need for this field in Inter-AS, but why would you need to check the SMET originating-ip for local bias? Ali> It is just inter-AS. -------------------------------------------------------------------------------- 4) Minor one: description of Maximum Response Time and Sequence number missing in section 7.3 and 7.3.1. Although both are roughly explained in section 4.2, the description of the fields and allowed values is missing in the section that describes the IGMP Leave synch route. Ali> We'll add it. Cheers, Ali -------------------------------------------------------------------------------- The below email captures the points I made during the adoption, but they are no longer valid anyway, so please, disregard. However the above points are the ones I think should be addressed now. Thank you! Jorge -----Original Message----- From: "Rabadan, Jorge (Nokia - US)" <[email protected]> Date: Thursday, February 9, 2017 at 8:30 AM To: Thomas Morin <[email protected]>, "[email protected]" <[email protected]> Cc: "[email protected]" <[email protected]> Subject: Re: [bess] Call for adoption: draft-sajassi-bess-evpn-igmp-mld-proxy-01 I support this document for WG adoption. Having said that, I made a few observations to the authors, and I believe they agreed to make some changes in the next revision. The main things that I believe should be reflected in the next rev after WG adoption are: 1- Simplified BGP route encoding I discussed with the authors that the Join and Leave synch behavior may have been achieved with a single route type, as opposed to the proposed two types (type 7 and 8). The authors believe it is better to keep both, which is ok, but: a) the route type 8 – IGMP leave synch route – should be simplified: the max response time and sequence number fields in the route introduce an unnecesary complexity and should be removed. b) Route type 8 should be optional since: i) It is actually not needed for IGMPv1 and 2) It is not needed either if a fast leave mechanism is used (see point 2). 2- Fast Leave addition to the draft There are quite a few igmp-snooping implementations in the market that support a “Fast Leave” mechanism. EVPN should incorporate/document this too. Implementations allow the use of "Fast Leave" when the IGMP host is directly connected to the PE/NVE and, only in that case is recommended. Fast Leave is a local administrative option on the PE, that, if enabled, allows the removal of a (x,G) state immediately after the reception of an IGMP Leave message for the (x,G). In the case of an ES AC, Fast Leave is only allowed in the case that a single IGMP host is multi-homed to the PEs in the ES. When Fast Leave is configured in an ES AC, the reception of an IGMP Leave message will remove the (x,G) state for the ES AC immediately and will trigger the withdrawal of the IGMP State Synch route. Assuming the remote PE is configured for "Fast Leave" too, the reception of the (x,G) route withdrawal for the ES will remove the (x,G) state completely. 3- Multicast Flags EC The Tunnel Type field looks not big enough for the different tunnel types that EVPN can use. I would recommend taking more space from the reserved bits and include all the allocated tunnel types in here: http://www.iana.org/assignments/bgp-parameters/bgp-parameters.xhtml#pmsi-tunnel-types Thank you. Jorge On 1/31/17, 3:58 PM, "BESS on behalf of Thomas Morin" <[email protected] on behalf of [email protected]> wrote: Hello working group, This email starts a two-week poll on adopting draft-sajassi-bess-evpn-igmp-mld-proxy-01 [1] as a working group item. Please send comments to the list and state if you support adoption or not (in the later case, please also state the reasons). This poll runs until **February 14th**. *Coincidentally*, we are also polling for knowledge of any IPR that applies to this draft, to ensure that IPR has been disclosed in compliance with IETF IPR rules (see RFCs 3979, 4879, 3669 and 5378 for more details). ==> *If* you are listed as a document author or contributor please respond to this email and indicate whether or not you are aware of any relevant IPR. The draft will not be adopted until a response has been received from each author and contributor. If you are not listed as an author or contributor, then please explicitly respond only if you are aware of any IPR that has not yet been disclosed in conformance with IETF rules. Thank you, Martin & Thomas bess chairs [1] https://datatracker.ietf.org/doc/draft-sajassi-bess-evpn-igmp-mld-proxy-01 _______________________________________________ BESS mailing list [email protected] https://www.ietf.org/mailman/listinfo/bess _______________________________________________ BESS mailing list [email protected] https://www.ietf.org/mailman/listinfo/bess
