FAQ-Why does traffic forwarding fail when the L3VPN label distribution method changes from per-route label to per-instance label

For S series switches (except S1700 switches), why does traffic forwarding fail when the L3VPN label distribution method changes from per-route label to per-instance label?

When this problem occurs, check whether a large number of BGP LSPs exist on the switch before the change of the label distribution method. (BGP LSPs occupy resources in the MPLS decapsulation table.)
In V200R003 and earlier versions, the per-route label distribution method is used for L3VPN. The number of entries in the MPLS decapsulation table of a switch is limited. Besides, the MPLS decapsulation table is a hash table and may encounter hash collisions. When there are a large number of local routes for an L3VPN, many entries in the MPLS decapsulation table are occupied. The per-instance label distribution method is recommended in this case.
When the label distribution method changes from per-route label to per-instance label, new BGP LSPs are generated in the VPN instance. If a large number of BGP LSPs exist on the switch before the change, the new BGP LSPs may cause hash collisions in the MPLS decapsulation table. As a result, traffic forwarding fails in the L3VPN.

Operation Notes:
Before changing the label distribution method to per-instance label on the switch, check the number of existing BGP LSPs. If there are a large number of BGP LSPs, withdraw local routes in the L3VPN to reduce the number of BGP LSPs, and then change the label distribution method.

Scroll to top