Transcript PPT Version
BGP/MPLS Traffic Blackhole Avoidance Proposal draft-asati-bgp-mpls-blackhole-avoidance-00 Rajiv Asati Raymond Zhang Tom Nadeau Azhar Sayeed IETF 68, March 21st 2007 Prague draft-asati-bgp-mpls-blackhole-avoidance-00.txt 1 Agenda Background / Motivation Problem Statement Solution Scope Next Steps draft-asati-bgp-mpls-blackhole-avoidance-00.txt 2 Background / Motivation MPLS data plane failure may occur due to The MPLS network pretends to have the reachability to the remote BGP prefixes, even during the (MPLS) data plane failure. LDP failure Label mismatch Forwarding entry corruption Misconfiguration.. The network continues to advertise the prefix reachability to the outside world. It is not optimal to attract the (VPN) customer traffic and blackhole it inside the MPLS network. Sub-optimal or Plain WRONG? draft-asati-bgp-mpls-blackhole-avoidance-00.txt 3 Background / Motivation Deployment scenario#1 – CE2 is dual-homed. The Site1->site2 traffic may get blackholed during the PE1->PE3 LSP failure. Without iBGP multipath at PE1, all site-to-site traffic get blackholed With iBGP multipath at PE1, some traffic get blackholed. This blackholing may happen independent of whether the LSP is setup in an “ordered” or “independent” mode, though the ordered mode may help to avoid the blackholing during LDP failure. MP-BGP PE1 selects the BGP path via PE3 PE1 PE2 P1 P3 LSP failure MPLS Backbone CE1 Site#1 PE1->PE3 LSP PE1->PE2 LSP Routing Protocol PE3 Despite the MPLS data plane failure, PE1 is unaware and CE1 continues to prefer PE1 to deliver the CE2-destined traffic. draft-asati-bgp-mpls-blackhole-avoidance-00.txt CE2 Site#2 4 Background / Motivation Deployment Scenario#2 – Single-homed CE2 site connected to CE1 via a backup link, and PE1->P2 LSP fails. Site1->site2 traffic may get blackholed. CE1 may not activate the backup path (via 2nd SP or ISP, or dial-up or p2p link etc.), since PE1 continues to advertise the reachability to the site#2. MP-BGP PE1 P1 LSP failure P3 PE2 CE2 MPLS Backbone CE1 Backup link Site#1 PE1->PE2 LSP Routing Protocol Site-to-site Backup link Despite the MPLS data plane failure, CE1 continues to prefer (dial-up/Internet/FR..) PE1 to deliver the CE2-destined traffic. draft-asati-bgp-mpls-blackhole-avoidance-00.txt Site#2 5 Problem Statement Labeled BGP prefixes (such as VPNv4) depend on the MPLS path to the NEXT_HOP BGP bestpath selection algorithm currently considers only IP reachability to the NEXT_HOP BGP is not aware of the MPLS reachability to the NEXT_HOP. draft-asati-bgp-mpls-blackhole-avoidance-00.txt 6 Solution BGP to be made aware of the MPLS reachability to the NEXT_HOP. The label availability doesn’t mean much. ‘BGP bestpath selection’ algorithm needs to include ‘MPLS reachability’ to the NEXT_HOP as an additional criterion. The MPLS reachability to the NEXT_HOP could be validated and recorded in the “LSP Health Database” (LHD). draft-asati-bgp-mpls-blackhole-avoidance-00.txt 7 Scope of the Draft Proposes a mechanism (within BGP) to consider the valid "MPLS path" to the NEXT_HOP of the BGP path, before qualifying that BGP path as the bestpath candidate. Valid MPLS path = Functional LSP (not just the label). Does not intend to Assume that ‘LSP failure’ always equals ‘broken LDP session’. Enforce any particular LSP validation technique and frequency Replace the MPLS FRR Suggest the techniques to keep the LHD* up-to-date Explain how the ‘LSP validation’ should be performed * LHD=LSP Health Database draft-asati-bgp-mpls-blackhole-avoidance-00.txt 8 Advantages After detecting the LSP failure to PE2, PE1 disqualifies the BGP path from PE2. After PE1 withdraws the route from CE1, CE1 can select the backup path (via 2nd SP, ISP, Dial-up etc), and restore the site-to-site connectivity. PE1 disqualifies the BGP path via PE2 and withdraws it from CE2 MP-BGP PE1 P1 LSP failure P3 PE2 MPLS Backbone CE2 CE1 Site#1 PE1->PE2 LSP Routing Protocol Backup link CE1 can reroute (dial-up/Internet/FR..) the traffic over the alternative path draft-asati-bgp-mpls-blackhole-avoidance-00.txt Site#2 9 Next Step WG to suggest… draft-asati-bgp-mpls-blackhole-avoidance-00.txt 10 draft-asati-bgp-mpls-blackhole-avoidance-00.txt THANK YOU! draft-asati-bgp-mpls-blackhole-avoidance-00.txt 11 Additional Slide#1 There are a lot of deployments that rely on out-of-band LSP health checks to detect the LSP failure Usage of internal or external toolkit to detect the LSP failure Such out-of-band mechanisms don’t do anything wrt MPLS VPN traffic blackholing draft-asati-bgp-mpls-blackhole-avoidance-00.txt 12