Comware

I have several 5930 (Commware 7 R26xx) with BGP vpnv4 address family peering, and I see that if trafic originating from a switch is indeed spread among the links, trafic onlyu transiting (thus I think the label routing) only uses one of the link.

dis bgp routing-table vpnv4 outlabel shows several outlabel per route (one per peer) but dis bgp routing-table vpnv4 inlabel shows only one.

What would be the best way to achieve load sharing between the different lionks ? (4 links between 2 switches, each with it's bgp peering, in IPv4 address family ecmp works fine with a balance ecmp 4)

Hi @Er_DA 

You can achieve load sharing between the different links by configuring AS_Path prepend OR local preference attribute on link for incoming routes.

Ok, I did not explain very weel : see that diagram

When trafic is injected in the MPLS network at PE1, using ebgp balance config, it uses both links. However, at MPLS routers, all traffic is sent on only one link. I'd like all links to be used. MPLS.

Setup is done with MP-BGP and the vpnv4/vpnv6 address families. I found mentions of additional-paths and vpn-route cross multipath , but in H3C commware docs, not on HPE ones

AS prepend and local preference are useful to get priority of trzfic, not using several equivalent (same cost) links.

Hello Er_DA,

Please use the following links in regards to your question:

Enabling ECMP VPN route redistribution (hpe.com)

Following is the link to the MPLS guide:

HPE FlexFabric 5940 & 5930 Switch Series MPLS Configuration Guide

Under following sections and chapters as indicated below you can find more information:

Overview > ECMP VPN route redistribution

and 

Configuring MPLS L3VPN > MPLS L3VPN configuration task list

and 

Configuring MPLS L3VPN > Enabling ECMP VPN route redistribution

Hope this helps!

Thanks for the Tip.

I tried it, but no change in the trafic pattern.

I also tried to add in the vpnv4 address family

peer <name> additional-paths send-receive
additional-paths select-best 2

Did not change anything either.

Obviously I tried reseting bgp after changong configuration, same result.