[j-nsp] VPLS issues

Richard A Steenbergen ras at e-gerbil.net
Fri Nov 30 11:41:19 EST 2012


Does anybody have any experience with forced LSP path selection for VPLS 
circuits? Long story short, when we fire up traffic on one particular 
VPLS instance, we're seeing SOME of the traffic it's carrying being 
blackholed. The pattern is one of certain IP or even TCP port pairs 
being blocked, and it seems to rotate over time, which screams "hashing 
across multiple LSPs where one of them is doing something bad, and it 
changes as the LSPs resignal over time" to me. To try and lock this 
down, I'm trying to force the VPLS traffic to route over a single LSP, 
in the usual manner with a forwarding-table export policy, and a very 
simple extended community regexp against the vrf-target community.

term VPLS {
    from community MATCH_VPLS;
    then {
        install-nexthop lsp-regex .*-SILVER.*;
        load-balance per-packet;
        accept;
    }
}

But it sure as hell doesn't look like it's narrowing the LSP selection:

ras at re0.router> show route forwarding-table family vpls table blah
Routing table: blah.vpls
VPLS:
Destination        Type RtRef Next hop           Type Index NhRef Netif
...
00:xx:xx:xx:xx:xx/48 user     0                  indr 1050634     5
                                                 idxd  3223     2
                   idx:1      xx.xx.142.132     Push 262153, Push 655412(top)  4543     1 xe-7/3/0.0
                   idx:1      xx.xx.142.62      Push 262153, Push 752660, Push 691439(top)  1315     1 xe-4/1/0.0
                   idx:2      xx.xx.142.132     Push 262153, Push 758372(top)  1923     1 xe-7/3/0.0
                   idx:2      xx.xx.142.62      Push 262153, Push 382341, Push 691439(top)  2541     1 xe-4/1/0.0
                   idx:3      xx.xx.142.132     Push 262153, Push 758372(top)  1923     1 xe-7/3/0.0
                   idx:3      xx.xx.142.62      Push 262153, Push 382341, Push 691439(top)  2541     1 xe-4/1/0.0
                   idx:4      xx.xx.142.30      Push 262153, Push 714676(top)  1500     1 xe-4/1/1.0
                   idx:4      xx.xx.142.62      Push 262153, Push 619458, Push 378636(top)  3864     1 xe-4/1/0.0
                   idx:xx     xx.xx.142.82      Push 262153, Push 601828(top)   989     1 xe-5/0/0.0
                   idx:xx     xx.xx.142.132     Push 262153, Push 684644(top)  3516     1 xe-7/3/0.0
                   idx:xx     xx.xx.142.62      Push 262153, Push 528898, Push 760875(top)  4766     1 xe-4/1/0.0
                   idx:xx     xx.xx.142.62      Push 262153, Push 792036, Push 691439(top)  3473     1 xe-4/1/0.0

Any ideas, about this or about troubleshooting the forwarding plane for 
VPLS in general? Other than that VPLS just sucks... :)

-- 
Richard A Steenbergen <ras at e-gerbil.net>       http://www.e-gerbil.net/ras
GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)


More information about the juniper-nsp mailing list