MPLS on top of our ip network with label filtering

[gnijs]

Hello,

I want to test/deploy MPLS on top of our current L3 network, but i do not want our current global routing table to be label-switched (which is the default behaviour).

This should be possible if i deploy new dedicated Loopbacks for MPLS connections + do ldp label filtering on the PEs to determine what is announced inside MPLS. I tried to simulate this in a lab, but for some reason, the routers always announces ALL loopbacks. I even used "no mpls ldp advertise-labels" (which should stop all advertisements, no ?), but i still see the loopback on the next hop:

PE1

int lo0

 desc Loopback for IP (OSPF)

 ip add 4.4.4.1 255.255.255.255

int lo100

 desc Loopback for MPLS

 ip add 100.100.100.1 255.255.255.255

int Gi1/0

 desc to P router

 ip add 1.1.1.1 255.255.255.0

 mpls ip

no mpls ldp advertise-labels

mpls ldp router-id Loopback100 force

on the P router (next hop):

P# show mpls forwarding-table

tag     tag or vc     prefix    outgoing int   next hop

20      Untagged   4.4.4.1/32                Gi1/0    1.1.1.1       <---- this one is still here, i do not want to announce 4.4.4.1, only 100.100.100.1

21      Untagged   100.100.100.1/32   Gi1/0    1.1.1.1

Could this be a bug ? running 12.4(10a) on 7200 router

regards,

Geert

[gnijs]

ok, i always seem to solve my own questions :-)

Key point to remember is that local label generation is based on the local routing table, not the local interface table. Since the local routing table also contains the remote routes, a router by default also generates labels for remote routes. My filter needs to be based on the local routing table, not the local interfaces. I need to adjust my access list or prefix list matching to the following:

ip prefix-list MPLS-LO permit 100.100.100.0/24 ge 32 le 32

conf t

 mpls ldp label

    allocate global prefix-list MPLS-LO

only the 100.100.100.x loopbacks are announced into ldp

sh mpls forwarding

Label

21    16       100.100.100.2/32

22     Pop    100.100.100.3/32

24     19       100.100.100.1/32

<end>

On my PEs i can now use MPBGP with these next hops to run BGP on top of my existing network without affecting my existing core routing...

See also: http://www.cisco.com/c/en/us/td/docs/ios/mpls/configuration/guide/convert/mp_ldp_book/mp_ldp_alloc_filter.html