[j-nsp] Junos MPLS question

Sorin CONSTANTINESCU consta at gmail.com
Mon Dec 12 03:23:51 EST 2005


On 12/12/05, Hannes Gredler <hannes at juniper.net> wrote:

Hi, Hannes.

> if the link-layer b/w R2,R3 support vrf-table-label
> then i'd configure vrf-table-lable on the egress
> VRF to make the load-balancing hash-calculation
> entirely dep. on IP header fields where you have
> more bits available for the hash-calculation (read: randomness)
> compared the MPLS/IP egress lookup that was recommended
> in a previous posting.
>

Thanks for the tip. I'll have the oppotrunity to test this soon.

> /hannes
>
> Sorin CONSTANTINESCU wrote:
> > Hi,
> >
> > I have a setup like this:
> >
> > R1 -> R2 -> R3 ->R4
> >
> > - R1,R2 and R3 are in an MPLS Domain. R4 is outside.
> > - between R3 and R4 are 4 links (load-balanced)
> > - R1 - Cisco
> > - R2, R3, R4 - Juniper
> >
> > The goal of this setup is to have traffic from R1 (inside VRF) to go
> > to R4 ballanced on the 4 links between R3 and R4.
> >
> > I have configured a VRF in R1 and R3. In R3 i have a static default
> > route to inet.0. Unfortunatelly, i cannot redistribute it via MBGP to
> > R1, so i use 4 fake routes with qualified-next-hop, which are
> > redistributed in BGP. In R1, i route 0/0 towards the 4 next-hops
> > received earlier via MBGP.
> >
> > All is Ok, but sometimes i hit a bug on the cisco router, and traffic
> > is no-more ballanced, and all traffic from R1 to R4 goes only on one
> > link (instead of four).
> >
> > I tried routing one prefix (10.255.255.69, let's say) via 4
> > qualified-next-hops (assiociated with the 4 links). On R1, i have:
> >
> >    10.255.255.69/32 193.226.x.y nolabel/265056
> >                              193.226.x.y nolabel/265056
> >
> > On the R3, i have:
> >
> > === cut here ===
> > adonay at R3> show route label 265056
> >
> > mpls.0: 86 destinations, 86 routes (86 active, 0 holddown, 0 hidden)
> > + = Active Route, - = Last Active, * = Both
> >
> > 265056             *[VPN/170] 3d 14:49:40
> >                       to x.x.x.x via so-0/1/0.0, Pop
> >                       to x.x.x.x via so-0/1/1.0, Pop
> >                     > to x.x.x.x via so-0/1/2.0, Pop
> >                       to x.x.x.x via so-0/1/3.0, Pop
> >
> > adonay at R3>
> > === and here ===
> >
> > Although it looks very good, when tested, traffic is not load-balanced
> > across the 4 links, but goes out one of them only.
> >
> >
> > Any help is greatly appreciated.
> >
> > Thanks.
> >
> > PS: I have opened a case with JTAC a few days ago, bot no luck yet.
> > --
> > Sorin CONSTANTINESCU
> > JNCIS-M, CCNP
> > consta at gmail.com
> >
> > _______________________________________________
> > juniper-nsp mailing list juniper-nsp at puck.nether.net
> > http://puck.nether.net/mailman/listinfo/juniper-nsp
>


--
Sorin CONSTANTINESCU
JNCIS-M, CCNP
consta at gmail.com



More information about the juniper-nsp mailing list