TRILL Working Group H. Zhai Internet-Draft ZTE Intended status: Standards Track T. Senevirathne Expires: June 13, 2013 Cisco Systems R. Perlman Intel Labs D. Eastlake 3rd M. Zhang Huawei F. Hu ZTE December 10, 2012 RBridge: Pseudo-Nickname draft-hu-trill-pseudonode-nickname-04 Abstract RBridges provide frame forwarding service to layer2 switches or end stations at the edge of TRILL campus. As defined in [RFC6325], when there are multiple RBridges attached to the same LAN segment, only one edge RBridge is allowed to be the frame forwarder of a specific VLAN in order to avoid potential frame duplication and loops in the TRILL campus. However, in some application scenarios, for example an end station is multi-homed to multiple RBridges needs to improve the resiliency and increase the available network bandwidth of the connection. This means all those RBriges attached to the end station can act as the frame forwarders of a specific VLAN. This kind of active-active connection violates the definition above. Frame duplication and forwarding loops may happen and it may cause the flip-flopping of the egress RBridge nickname associated to the MAC address of such an end station in remote RBridges' forwarding tables. The Reverse Path Forwarding Check does not work as well, which has been addressed in [CMT]. This document proposes the concept of Virtual RBridge, along with the pseudo-nickname configuration for this Virtual RBridge, to address the above problems in accompany with [CMT]. Status of this Memo This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet- Drafts is at http://datatracker.ietf.org/drafts/current/. Zhai, et al. Expires June 13, 2013 [Page 1] Internet-Draft Pseudo-Nickname December 2012 Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." This Internet-Draft will expire on June 13, 2013. Copyright Notice Copyright (c) 2012 IETF Trust and the persons identified as the document authors. All rights reserved. This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License. Zhai, et al. Expires June 13, 2013 [Page 2] Internet-Draft Pseudo-Nickname December 2012 Table of Contents 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.1. Terminology and Acronyms . . . . . . . . . . . . . . . . . 5 1.2. Contributors . . . . . . . . . . . . . . . . . . . . . . . 5 2. Problem Statement . . . . . . . . . . . . . . . . . . . . . . 5 2.1. Appointed Forwarders on Shared Links . . . . . . . . . . . 6 2.2. Multi-homing and Link Aggregation to TRILL Network . . . . 6 3. Concept of Virtual RBridge and Pseudo-nickname . . . . . . . . 7 3.1. VLAN-x Appointed Forwarder for member interfaces in RBv . 7 3.2. Announcing Pseudo-Nickname of RBv . . . . . . . . . . . . 8 4. Distribution Trees for Member RBridges in RBv . . . . . . . . 8 4.1. Trees for Native Frames Ingressing . . . . . . . . . . . . 9 4.2. Designated Forwarder for traffic received on Distribution trees . . . . . . . . . . . . . . . . . . . . 9 5. Frame Processing . . . . . . . . . . . . . . . . . . . . . . . 11 5.1. Native Frames Ingressing . . . . . . . . . . . . . . . . . 12 5.2. TRILL Data Frames Egressing . . . . . . . . . . . . . . . 12 5.2.1. Unicast TRILL Data Frames . . . . . . . . . . . . . . 12 5.2.2. Multi-Destination TRILL Data Frames . . . . . . . . . 13 6. Member Link Failure in RBv . . . . . . . . . . . . . . . . . . 14 6.1. Link Protection for Unicast Frame Egressing . . . . . . . 14 6.2. Link Protection for Multi-destination Frame Egressing . . 15 7. Extensions to ESADI . . . . . . . . . . . . . . . . . . . . . 15 8. OAM Frames . . . . . . . . . . . . . . . . . . . . . . . . . . 17 9. Configuration Consistency . . . . . . . . . . . . . . . . . . 17 10. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 17 11. Security Considerations . . . . . . . . . . . . . . . . . . . 17 12. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 17 13. Normative References . . . . . . . . . . . . . . . . . . . . . 18 Appendix A. Rationale for MAC Sharing among Member RBridges . . . 18 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 20 Zhai, et al. Expires June 13, 2013 [Page 3] Internet-Draft Pseudo-Nickname December 2012 1. Introduction The IETF TRILL protocol [RFC6325] provides optimal pair-wise data frame forwarding without configuration, safe forwarding even during periods of temporary loops, and support for multi-pathing of both unicast and multicast traffic. TRILL accomplishes this by using [IS-IS] [RFC1195] link state routing and encapsulating traffic using a header that includes a hop count. The design supports VLANs and optimization of the distribution of multi-destination frames based on VLANs and IP derived multicast groups. Devices that implement TRILL are called RBridges. In TRILL protocol, RBridges are identified by nicknames (16-bits). At the edge of TRILL network, some RBridges connect to legacy networks on one side and connect to the TRILL network on the other side. These RBridges are called edge RBridges. For the connectivity between the two types of network, edge RBridges provide frame forwarding service to end stations located in legacy networks. When receiving a native frame from such a local end station S, the service edge RBridge RB1 encapsulates the frame in a TRILL header, addressing the packet to RBridge RBx to which the destination end station D is attached. The TRILL header contains an "ingress RBridge nickname" field (filled with RB1's nickname), an "egress RBridge nickname" field (filled with RBx's nickname), and a hop count. On receiving such a frame, RBx removes the TRILL header and forwards it in native form to D. Meanwhile, based on the de-capsulation of that frame, RBx learns the { ingress RBridge nickname, source MAC address, VLAN ID } triplet. Edge RBridges maintain such triplets in their forwarding tables for the future forwarding of native frames. In TRILL protocol, RBridges are identified by nicknames (16-bits). At the edge of TRILL network, some RBridges connect to legacy networks on one side and connect to the TRILL network on the other side. These RBridges are called edge RBridges. For the connectivity between the two types of network, edge RBridges provide frame forwarding service to end stations located in legacy networks. When receiving a native frame from such a local end station S, the service edge RBridge RB1 encapsulates the frame in a TRILL header, addressing the packet to RBridge RBx to which the destination end station D is attached. The TRILL header contains an "ingress RBridge nickname" field (filled with RB1's nickname), an "egress RBridge nickname" field (filled with RBx's nickname), and a hop count. On receiving such a frame, RBx removes the TRILL header and forwards it in native form to D. Meanwhile, based on the de-capsulation of that frame, RBx learns the { ingress RBridge nickname, source MAC address, VLAN ID } triplet. Edge RBridges maintain such triplets in their forwarding tables for the future forwarding of native frames. Zhai, et al. Expires June 13, 2013 [Page 4] Internet-Draft Pseudo-Nickname December 2012 In this document, the concept of Virtual RBridge group, together with its Pseudo-nickname, is introduced to address the rest of above issues. For a member RBridge of such a group, it uses the pseudo- nickname, instead of its own nickname, as the ingress RBridge nickname when ingressing frames into TRILL campus. So, in such a RBridge Group, even if there are more than one RBridge providing frame forwarding service for an end station or the service RBridge changes over from one to another member RBridge in same group, the ingress RBridge nickname associated to this end station needs not be unchanged in remote RBridges' forwarding tables. This document is organized as follows: Section 2 describes why virtual RBridge along with its pseudo-nickname is needed. Section 3 gives the concept of virtual RBridge. Section 4 describes distribution trees allocation across member RBridges in a RBv. Section 5 covers the processing of transit frames in consideration of the pseudo-nickname. Familiarity with [RFC6325] is assumed in this document. 1.1. Terminology and Acronyms This document uses the acronyms defined in [RFC6325] and the following additional acronym: AF - Appointed Forwarder The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in [RFC2119]. When used in lower case, these words convey their typical use in common language, and are not to be interpreted as described in [RFC2119]. 1.2. Contributors We would like to thank Mingjiang Chen for his contributions to this document. Additionally, we would like to thank Erik Nordmark, Les Ginsberg, Ayan Banerjee, Dinesh Dutt, Anoop Ghanwani, Janardhanan Pathang, and Jon Hudson for their good questions and comments. 2. Problem Statement Zhai, et al. Expires June 13, 2013 [Page 5] Internet-Draft Pseudo-Nickname December 2012 2.1. Appointed Forwarders on Shared Links Even there are multiple RBridges on a shared link, together with end stations, only one RBridge is allowed to provide frame forwarding services in VLAN-x to the end stations to avoid possible frame duplication or loops in TRILL campus. The service RBridge is called VLAN-x Appointed Forwarder (AF). However, AF for any set of VLANs on a shared link may change over from one RBridge to another, due to network dynamics such as failures and configuration changes. RBridges rely on LSPs to propagate these network dynamics. However, the propagation is time consuming and the network may take a considerable long time to converge. Before the network converges, remote RBridges may continue to forward traffic to the previous AF and the traffic is dropped at the previous egress RBridge, causing traffic disruption. 2.2. Multi-homing and Link Aggregation to TRILL Network In order to improve the reliability of connection to TRILL network, multi-homing technique may be employed by a legacy device which can be a switch or end host. Take Figure 1 as an example, switch SW1 multi-homes to TRILL network by connecting to RB1 and RB2 with respective links. Then the end station S1 can continue to get frame forwarding service from TRILL network even if one of its up-links (e.g., SW1-RB1) fails. .......................................... : TRILL Network : :+-----+ /\-/\-/\-/\-/\ : +-----| RB1 |-----/ \ : | :+-----+ / \ : +---+ : | / Transit \ +-----+ : S1 o--|SW1| +---+ < RBridges >---| RBx |---o Sx +---+ | : \ Campus / +-----+ : | | :+-----+ \ / : +-----| RB2 |-----\ / : | :+-----+ \/\-/\-/\-/\-/ : | : | : S2 o----+ : .......................................... Figure 1 Multi-homing to TRILL Network SW1 may treat the two links as a LAG (Link Aggregation Group) bundle, so that the two links form active-active load sharing model instead of previous active-standby model. That is to say, in Figure 1, two Zhai, et al. Expires June 13, 2013 [Page 6] Internet-Draft Pseudo-Nickname December 2012 RBridges (i.e., RB1 and RB2) provides frame forwarding service to S1 simultaneously in a VLAN. As stated previously, that simultaneous frame forwarding may result in frame duplication, loops and the flip- flopping of the ingress RBridge associated to the MAC of S1 in remote RBridges' (e.g., RBx) forwarding tables. The flip-flopping in turn causes packet disorder in reverse traffic and worsens the traffic disruption. Therefore, the concept of Virtual RBridge, together with its nickname, is introduced in the following section to fix these issues. 3. Concept of Virtual RBridge and Pseudo-nickname A Virtual RBridge (RBv) represents a group of different end station service ports on different edge RBridges. After joining RBv, such an RBridge port is called a member port of RBv, and such an RBridge becomes a member RBridge of RBv. An RBv is identified by its virtual nickname in TRILL campus, and this nickname is also referred to as pseudo-nickname in this document. After joining an RBv, a member RBridge will announce its connection to RBv by including the information of that RBv, e.g., the pseudo- nickname of RBv, in its self-originated LSP. From such LSPs, other RBridges that are not members of the RBv believe those member RBridges are connected to RBv. When a native frame from an end station S1 is received from such a port, the member RBridge encapsulates the frame with the RBv's nickname, instead of its own nickname, as the ingress nickname. When the destination RBridges receive and de-capsulate this frame, they will learn that S1 is reachable through RBv. NOTE: An RBridge port can join at most one RBv at any time, but different ports on the same RBridge can join the same RBv or different RBvs. Furthermore, for a member RBridge, it MUST move out of an RBv and clear the RBv's information from its self-originated LSPs when it loses all of its member ports of the RBv, due to port failure, configuration, etc. 3.1. VLAN-x Appointed Forwarder for member interfaces in RBv If member RBridges in RBv cannot see each others' Hellos on their member ports (e.g., in the LAG scenario), then each RBridge becomes Designated RBridge (DRB) for that port and appoints itself as AF for all VLANs. On the other hand, if they can see each others' Hellos on the member ports in RBv (e.g., in the shared link scenario), the TRILL Hello Zhai, et al. Expires June 13, 2013 [Page 7] Internet-Draft Pseudo-Nickname December 2012 protocol in [RFC6325] is used for DRB election and for VLAN-x AFs appointment on those ports. Then the DRB appoints different member ports as AFs for different sets of VLANs. By using the AF framework specified in [RFC6325], a unified framework of RBv for both LAG and shared LAN edge connectivity is provided in this document. It also allows o Detection and protection against mis-configuration at the edge, e.g., on the device SW1 the two interfaces are not configured as LAG then RB1 and RB2 work in an unexpected active-standby mode rather than expected active-active mode for S1 or o Avoidance of loops in the event that S1 and S2 were connected by a native Ethernet Link. In this event, RB1's Hellos originated on link RB1-SW1 will be forwarded by S1 through the Ethernet Link to S2 then received by RB2, and vice versa. Therefore, RB1 and RB2 work in an active-standby mode for S1 (or S2) in a VLAN to avoid potential forwarding loops. 3.2. Announcing Pseudo-Nickname of RBv Each member RBridge advertises the RBv's pseudo-nickname using the nickname sub-TLV [rfc6326bis], along with its regular nickname(s), in its LSPs. For a member RBridge, when its last member port is disconnected to RBv, it MUST leave from RBv and clear RBv's pseudo- nickname from its update LSPs. RBv's pseudo-nickname is ignored when determining the distribution tree root for the campus. The tree root priority of RBv's nickname SHOULD be set to 0, and this nickname SHOULD NOT be listed in the s nicknames by the RBridge holding the highest priority tree root nickname. 4. Distribution Trees for Member RBridges in RBv In TRILL, RBridges use distribution trees to forward multi- destination frames. To guarantee member RBridges of an RBv group to ingress (i.e., northbound forwarding) and egress (i.e., southbound forwarding) multi-destination frames properly, and to provide protection in case of member link failure as well, this document proposes to allocate the trees among member RBridges in northbound and southbound directions respectively. Section 4.1 describes the allocation of trees for the ingressing of multi-destination frames, while Section 4.2 covers the allocation of trees for the egressing of multi-destination frames. Then for each member RBridge, frames received from each direction can be forwarded based on the tree Zhai, et al. Expires June 13, 2013 [Page 8] Internet-Draft Pseudo-Nickname December 2012 allocation. 4.1. Trees for Native Frames Ingressing In the TRILL header of the multi-destination frames, the ingress nickname identifies the ingress RBridge and the egress nickname specifies the root of the chosen distribution tree. After receiving a multi-destination TRILL data frame, RBn performs Reverse Path Forwarding (RPF) check on the multi-destination frame to avoid temporary multicast loops during topology changes. RPF specifies that a multi-destination TRILL data frame ingressed by an RBridge and forwarded on a distribution tree can only be received by RBn on an expected port. If the frame is received not from that port, it MUST be dropped. However, member RBridges use RBv's pseudo-nickname other than their own nicknames as the ingress nickname when they forward northbound frames regardless unicast or non-unicast frames. Therefore, when these TRILL data frames arrive at RBn, they will be treated as frames ingressed by the same RBridge, i.e., RBv. If they are multi- destination frames and the same distribution tree is chosen by different member RBridges to forward these frames, they may travel on the tree and arrive at RBn on different ports. Then the RPF check is violated, and some of the frames reaching the RBridge on unexpected ports will be dropped by RBn. [CMT] proposes to assign different distribution trees for each member RBridge to fix the above RPF check issue, and makes use of the Affinity sub-TLV defined in [rfc6326bis] to achieve this kind of assignment. This document supposes the approach proposed in [CMT] is supported by member RBridges of RBv. When a member RBridge joins in or leaves from a virtual RBridge group, the assignment of distribution trees may change. That change beyond the scope of this document. 4.2. Designated Forwarder for traffic received on Distribution trees Methodology explained above addresses the association of distribution trees to RBx on behalf of RBv for native traffic received from the edge. However, remote RBridge RBn who is unaware of the RBv association may choose to forward traffic along any of the available distribution trees. Only one member RBridge RBx of the virtual RBridge is allowed to de-capsulate the TRILL frames and forwards them as native frames. In this document, this member RBridge is called Zhai, et al. Expires June 13, 2013 [Page 9] Internet-Draft Pseudo-Nickname December 2012 the Designated Forwarder. It is important to note that RBv represents multiple, multi-homed devices connected to the same RBv. Some of the devices may have operational links to all member RBridges of RBv and some may have operational links only to a subset of all member RBridges. The objective of this proposal is to identify a member RBridge RBx per distribution tree and end-device (LAG), that will de-capsulate and forward traffic received from a specific distribution tree(s) to specific end-device that multi-homes to the RBv. We propose to represent each of the end devices *multi-homed* to RBv by a unique identifier (LAG-ID) within the RBv group. Each of the RBx advertises { RBv, LAG-ID } using ESADI framework. Please see Section 7 for ESADI-TLV. Each RBx, in turn, performs election of Designated Forwarder for each LAG-ID per distribution tree "t". ------------------ / \ | | | TRILL Campus | | | \ / --------------------- | | | | | +--------+ | | | +------+ +------+ +------+ | RB1 | | RB2 | | RB3 | o|oooooo|oo|oooooo|oooooo|oooooo|o o +------+ +------+ +------+ o o | | | | | o o | | +----+ | | o o | Virtual|RBridge|(RBv)| | o ooo|oooooooo|ooooooo|ooooo|ooo|ooo | | +-----|-----+ | +------+ | | +-------+ | (| | |)<--LAG1 (| |)<--LAG2 +-------+ +-------+ | CE1 | | CE2 | +-------+ +-------+ Figure 2 RBv with Multiple Multi-homed Devices Assume RB(i) are the member RBridges of RBv that advertise Zhai, et al. Expires June 13, 2013 [Page 10] Internet-Draft Pseudo-Nickname December 2012 association of LAG-ID(j). Suppose there are T(m) distribution trees in the network. Step 1: Sort the RB(i) set in ascending order based on their system-id. Step 2: Sort distribution tree T(m) in ascending order based on their tree-number. Step 3: Allocate the sorted list in Step 2 in round-robin manner to the sorted RBridge list in step 1. Step 4: Repeat Step 1-3 for all of the LAG-ID associated with RBv. Through the above steps, within a specific LAG, each RBx is assigned with a set of distribution trees, and this RBx will become Designated Forwarder for these trees. RBx MUST only forward traffic (received on these trees) to member links on which it is the Designated Forwarder. For example, in Figure 2, CE1 multi-homes to RB1, RB2 and RB3, and CE2 multi-homes to RB2 and RB3 in the same RBv; CE1 and CE2 are represented by LAG1 and LAG2 respectively. Assume RB1 < RB2 < RB3 based on their system id, and there are five trees available in TRILL campus, say T1 < T2 < T3 < T4 < T5 based on their tree-number. After the election is performed in LAG1, T1 and T4 is allocated to RB1, T2 and T5 are allocated to RB2, and T3 is assigned to RB3. In other words, RB1 is the Designated Forwarder for LAG1 on T1 and T4, RB2 is the Designated Forwarder for LAG1 on T2 and T5 and RB3 is the Designated Forwarder for LAG1 on T3. After the election is performed for LAG2, T1, T3 and T5 are allocated to RB2, T2 and T4 are allocated to RB3. However, since only one member RBridge of RBv is elected as AF for e set of VLANs on a shared link and responsible to egress multi- destination TRILL data frames to this link, each member RBridge MUST assign itself as the Designated Forwarder for all the distribution trees to guarantee multi-destination frames on any tree will be egressed onto this link by an member RBridge (please refer Section 5.2.2 for more details). 5. Frame Processing Zhai, et al. Expires June 13, 2013 [Page 11] Internet-Draft Pseudo-Nickname December 2012 5.1. Native Frames Ingressing When RB1 receives a native frame on one of its member ports of RBv, it uses the pseudo-nickname of RBv, instead of its own nickname, as the ingress nickname, if it is the appointed forwarder for the VLAN of the frame on that port. If the frame is not received on a member port, RB1 MUST NOT use RBv's pseudo-nickname as ingress nickname for TRILL-encapsulation. If the frame is ingressed as a multi-destination TRILL data frame, RB1 can only choose one of its assigned distribution trees to forward the TRILL-encapsulated frame [CMT]. Otherwise, the multi-destination TRILL data frame will fail the RPF check on remote RBridges and suffer unexpected packet loss. Source MAC address learning is performed during the frame de- capsulation as described in [RFC6325]. The learned MAC address SHOULD be shared with other member RBridges within the same RBv group (See Appendix A for more details for the information sharing). 5.2. TRILL Data Frames Egressing This section describes egress processing of the received TRILL data frames on a member RBridge (say RBn) in the virtual RBridge group. Section 5.2.1 describes unicast TRILL data frames egress processing and Section 5.2.2 specifies the multi-destination TRILL data frames egressing. 5.2.1. Unicast TRILL Data Frames When receiving a unicast TRILL data frame, RBn checks the egress nickname in the TRILL header of the frame. If the egress nickname is one of RBn's own nicknames, the frame is processed as defined in in [RFC6325]. If the egress nickname is RBv's pseudo-nickname and RBn is a member RBridge of RBv, RBn is responsible to learn the source MAC address. If the learned { Inner.MacSA, Inner.VLAN ID, ingress nickname } triplet is a new one or it updates a previously learned one, this triplet SHOULD be shared with other member RBridges within the RBv (See Appendix A for more details for the triplet sharing). Then the frame is de-capsulated to its native form. The Inner.MacDA and Inner.VLAN ID are looked up in RBn's local forwarding address cache, and one of the three following cases occurs: Zhai, et al. Expires June 13, 2013 [Page 12] Internet-Draft Pseudo-Nickname December 2012 o If the destination end station identified by the Inner.MacDA and Inner.VLAN ID is on a local link to RBv, this frame is egressed onto that link *regardless of whether RBn is the Inner.VLAN AF on this link*. o Else if RBn can reach the destination through another member RBridge RBk, it re-encapsulates the native frame into a unicast TRILL data frame and sends it to RBk. RBn uses RBk's own nickname, instead of RBv's pseudo-nickname as the egress nickname for the re-encapsulation, and remains the ingress nickname unchanged. If the hop count value of the frame is too small for the frame to reach RBk safely, RBn SHOULD increase that value properly in doing the re-encapsulation. [NOTE: When receiving that re-encapsulated TRILL frame, as the egress nickname of the frame is RBk's own nickname rather than the RBv's pseudo-nickname, RBk will process it as Section 4.6.2.4 in [RFC6325]., and will not re-forward it to another RBridge.] o Else, RBn does not know how to reach the destination; it sends the native frame out of all its member ports of RBv on which it is appointed forwarders for the Inner.VLAN. 5.2.2. Multi-Destination TRILL Data Frames If RBn is the AF for the Inner.VLAN, the source MAC address is learned. If the learned { Inner.MacSA, Inner.VLAN ID, ingress nickname } triplet is a new one or updates a previously learned one, this triplet SHOULD be shared among the members RBridges within the virtual RBridge group (See Appendix A for more details for the triplet sharing). Then a copy of the frame is de-capsulated into its native form. Before the native frame is sent out of the ports on which RBn is appointed forwarder for the Inner.VLAN, the following extra check is performed for each member port of RBv: o Frames MUST only be forwarded out on member ports of RBv where RBn is the Designated Forwarder for the Tree Tn on which the frame was received. Zhai, et al. Expires June 13, 2013 [Page 13] Internet-Draft Pseudo-Nickname December 2012 6. Member Link Failure in RBv ------------------ / \ | | | TRILL Campus | | | \ / --------------------- | | | | | +--------+ | | | +------+ +------+ +------+ | RB1 | | RB2 | | RB3 | o|oooooo|oo|oooooo|oooooo|oooooo|o o +------+ +------+ +------+ o o | | | | | o o | | +----+ | | o o | Virtual|RBridge|(RBv)| | o ooo|oooooooo|ooooooo|ooooo|ooo|ooo | \|/ | +-----|-----+ | +- B --+ | | +-------+ | /|\ (| | |)<--LAG1 (| |)<--LAG2 +-------+ +-------+ | CE1 | | CE2 | +-------+ +-------+ B - Failed Link or Link bundle Figure 3 Member Link Failure in LAG1 As shown in Figure 3, suppose the link RB1-CE1 fails. Both unicast frames and multi-destination frames cannot be sent from RB1 to CE1. Section 6.1 discusses the failure protection for unicast frames receiving, and Section 6.2 describes the failure protection for multi-destination frames. 6.1. Link Protection for Unicast Frame Egressing When the link CE1-RB1 fails, RB1 loses the connection to the local link to CE1. The MAC entry through the failed link to CE1 is removed from RB1's local forwarding table immediately. Another MAC entry through another member RBridge (say RB2) that has local link to CE1 is installed into RB1's forwarding table only if RB1 is still a member RBridge of RBv. Then when TRILL data frames to CE1 is delivered to RB1, it can be re-encapsulated (ingress nickname remains unchanged and egress nickname is replaced by RB2's nickname) by RB1 Zhai, et al. Expires June 13, 2013 [Page 14] Internet-Draft Pseudo-Nickname December 2012 and forwarded based on the above installed MAC entry. The member RBridge who receives the redirected frames will egress them to CE1. When the failure recovers, RB1 will be aware that it can reach CE1 by observing CE1's native frames. Then RB1 installs the MAC entry for link RB1-CE1, which restores the link CE1-RB1. 6.2. Link Protection for Multi-destination Frame Egressing When RBx loses all of its links to a given edge device (represented by a unique LAG-ID), it MUST advertise membership withdrawal. This in turn triggers other RBridges to re-calculate the Designated Forwarder allocation for the device. For example, in Figure 3, let's assume RB1 is the Designated Forwarder on T1 and T4 for CE1 (identified as LAG1), and is egressing multi-destination traffic from T1 to CE1 before the failure. When that failure occurs, RB1 loses its only local link to CE1 and advertises to other RBridges its membership withdrawal for LAG1. This triggers RB2 and RB3 to re-calculate the Designated Forwarder allocation for LAG1. After the re-calculation, trees T1, T3 and T5 are allocated to RB2, and T2 and T4 to RB3. That is to say, RB2 becomes the Designated Forwarder on the 3 trees and RB3 becomes the Designated Forwarder on the 2 trees for LAG1. Then RB2 takes the place of RB1 to egress the traffic to CE1. Similarly, when the first attachment link between RBx and an edge device becomes operational, RBx joins in the virtual RBridge group and advertises membership adding. This in turn triggers other RBridges as well as itself to re-calculate the Designated Forwarder allocation for the device. 7. Extensions to ESADI We propose to use sub-TLV within ESADI [ESADI] to advertise following 3 different scenarios. o Membership Add o Membership Withdrawal o Membership Refresh Zhai, et al. Expires June 13, 2013 [Page 15] Internet-Draft Pseudo-Nickname December 2012 +-+-+-+-+-+-+-+-+ | Type= LM | (1 byte) +-+-+-+-+-+-+-+-+ | Length | (1 byte) +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | RBv Nickname | (2 bytes) +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | RESV |OC | (1 byte) +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | LAG-ID (1) | (2 bytes) +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ . . +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | LAG-ID (n) | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ Figure 4 Edge Membership advertisement sub-TLV o LM (1 byte): Defines the type of Edge Membership sub-TLV. o Length (1 byte): Defines the length of this sub-TLV which should be greater than 3. o RBv Nickname (2 bytes): 2 byte nickname of the RBv assigned to the edge group. o RESV (6 bits): Transmitted as zero and ignored on receipt. o OC (2 bits): Define the operation code. * 00: Add (LAG-IDs in this sub-TLV are new and need to be reconsidered for Designated Forwarder election). * 01: Withdrawal (LAG-IDs in this sub-TLV do not have an active links from the announcing RBridge for RBv and Designated Forwarder election MUST be recalculated). * 10: Refresh (LAG-IDs in this sub-TLV are being refreshed and no state change from the perspective of the announcing RBridge). * 11: Reserved and currently unused. o LAG-ID (2 bytes): an unsigned positive integer that uniquely identifies an end device multi-homed to the RBv. When receiving such a sub-TLV, if the RBridge has no membership for the listed LAGs in the RBv, it ignores the sub-TLV. If it has the Zhai, et al. Expires June 13, 2013 [Page 16] Internet-Draft Pseudo-Nickname December 2012 membership, receiving such a sub-TLV where the operation code is 00 or 01 will triggers it to re-calculate the Designated Forwarder on each tree for the listed LAGs. 8. OAM Frames Attention must be paid when generating the OAM frames. When an OAM frame is generated with the ingress nickname of RBv, the originator RBridge's nickname MUST be included in the OAM message to ensure the response is returned to the originating member of the RBv group. 9. Configuration Consistency It is important that VLAN membership of member ports of end switch SW1 is consistent across all of the member ports in the point-point scenario. Any inconsistencies in VLAN membership may result in packet loss or non-shortest paths. Take Figure 1 for example, suppose RB1 configures VLAN1 and VLAN2 for the link SW1-RB1, while RB2 only configures VLAN1 for the SW1-RB2 link. Both RB1 and RB2 use the same ingress nickname RBv for all frames originating from S1. Hence, a remote RBridge RBx will learn that MAC addresses from S1 on VLAN2 are originating from RBv. As a result, on the returning path, RBx may deliver VLAN2 traffic to RB2. However, RB2 does not have VLAN2 configured on SW1-RB2 link and hence the frame may be dropped or has to be redirected to RB1 if RB2 knows RB1 can reach S1 in VLAN2. 10. IANA Considerations TBD. 11. Security Considerations TBD. 12. Acknowledgements We would like to thank Mingjiang Chen for his contributions to this document. Additionally, we would like to thank Erik Nordmark, Les Ginsberg, Ayan Banerjee, Dinesh Dutt, Anoop Ghanwani, Janardhanan Pathang, and Jon Hudson for their good questions and comments. Zhai, et al. Expires June 13, 2013 [Page 17] Internet-Draft Pseudo-Nickname December 2012 13. Normative References [CMT] Senevirathne, T., Pathangi, J., and J. Hudson, "Coordinated Multicast Trees (CMT)for TRILL", draft-ietf-trill-cmt-01.txt Work in Progress, November 2012. [ESADI] Zhai, H., Hu, F., Perlman, R., and D. Eastlake, "TRILL (Transparent Interconnection of Lots of Links): The ESADI (End Station Address Distribution Information) Protocol", draft-ietf-trill-esadi-01.txt Work in Progress, October 2012. [RFC1195] Callon, R., "Use of OSI IS-IS for routing in TCP/IP and dual environments", RFC 1195, December 1990. [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. [RFC6325] Perlman, R., Eastlake, D., Dutt, D., Gai, S., and A. Ghanwani, "Routing Bridges (RBridges): Base Protocol Specification", RFC 6325, July 2011. [rfc6326bis] Eastlake 3rd, D., Banerjee, A., Ghanwani, A., and R. Perlman, "Transparent Interconnection of Lots of Links (TRILL) Use of IS-IS", draft-ietf-isis-rfc6326bis-00.txt Work in Progress, October 2012. Appendix A. Rationale for MAC Sharing among Member RBridges With the introduction of virtual RBridge, MAC flip-flopping problem in LAN or LAG is resolved. However, in order to forward traffic effectively, member RBridges should share some of their learned MAC addresses with each other. Zhai, et al. Expires June 13, 2013 [Page 18] Internet-Draft Pseudo-Nickname December 2012 ........................................... ooooo TRILL Network : o : +-----+ /\-/\-/\-/\-/\ : +---o----| RB1 |----/ \ : | o : +-----+ / \ : +-----+o : o | / Transit \ +-----+ : S1 o--| SW1 |o RBv o | < RBridges >---| RBx |---o Sx +-----+o : o | \ Campus / +-----+ : | o : +-----+ \ / : +---o----| RB2 |----\ / : o +-----+ \/\-/\-/\-/\-/ : ooooo : ........................................... Figure 5 RBv in LAG scenario Take Figure 5 as an example, the VLAN-x native frames from S1 to Sx will enter TRILL campus via one member RBridge of the RBv (say RB1). RB1 learns the location of S1 in VLAN-x. However, RBx may deliver the reverse traffic to RB2 if it thinks the shortest path to RBv is through RB2. If RB2 has not learned the location of S1 in VLAN-x from the MAC sharing, RB2 has to transmit the reverse traffic to S1 as unknown unicast. Thus, the learned MAC addresses of attached end stations on one member RBridge SHOULD be shared with the rest member RBridges in the same RBv. With these information shared, when RB2 receives reverse frames, it can determine how to forward them to S1. For example, it can redirect them to RB1 if link RB2-S1 fails. Since RBx always delivers the reverse traffic to RBv via RB2, RB2 egresses the traffic and learns the location of Sx. But RB1 will not know where Sx is, if RB2 does not share this information with RB1. As a result, RB1 has to treat the traffic from S1 to Sx as traffic with unknown destination and flood it in TRILL, which adds additional forwarding burden on TRILL network. Therefore, in addition to local attached end station MAC addresses, the learned remote MAC addresses should also be shared among all member RBridges of an RBv. With such information shared, RB1 can treat the traffic to Sx as known destination traffic and unicast it to RBx. The design for above MAC sharing is currently beyond the scope of this document. Zhai, et al. Expires June 13, 2013 [Page 19] Internet-Draft Pseudo-Nickname December 2012 Authors' Addresses Hongjun Zhai ZTE 68 Zijinghua Road, Yuhuatai District Nanjing, Jiangsu 210012 China Phone: +86 25 52877345 Email: zhai.hongjun@zte.com.cn Tissa Senevirathne Cisco Systems 375 East Tasman Drive San Jose, CA 95134 USA Phone: +1-408-853-2291 Email: tsenevir@cisco.com Radia Perlman Intel Labs 2200 Mission College Blvd Santa Clara, CA 95054-1549 USA Phone: +1-408-765-8080 Email: Radia@alum.mit.edu Donald Eastlake 3rd Huawei 155 Beaver Street Milford, MA 01757 USA Phone: +1-508-333-2270 Email: d3e3e3@gmail.com Zhai, et al. Expires June 13, 2013 [Page 20] Internet-Draft Pseudo-Nickname December 2012 Mingui Zhang Huawei Huawei Building, No.156 Beiqing Rd. Beijing, Beijing 100095 China Email: zhangmingui@huawei.com Fangwei Hu ZTE 889 Bibo Road, Pudong District Shanghai, Shanghai 201203 China Phone: +86 21 68896273 Email: hu.fangwei@zte.com.cn Zhai, et al. Expires June 13, 2013 [Page 21]