MPLS unicast IP forwarding

[Ali Norouzi]

I've MPLS unicast IP forwarding applicationn on Cisco router. Now I want to have same on Juniper M series routers. The Juniper is PE in my network and OSPF and LDP is running on it. It learns and advertises the lables correctly. Learned routes are even installed on inet.3 routing table. But the problem is that there is no lable in inet.0 routing table. So when ip traffic ingresses the router, it forwards it to next-hop without lable. Do you know how I can forward those incoming packet with MPLS lable to next-hop?

Thank you,

[Reza Sharifi]

Hi Ali,

That is the correct behavior:

By default, the only prefix installed in the inet.3 routing table is the /32 address associated with the LSP endpoint. You can add additional prefixes to the inet.3 table using the install keyword or enabling TE on your IGP and MPLS. Once that is done, then your prefixes will available for BGP next hop resolution

If you are running for example OSPF as your IGP, you need to enable

set traffic-engineering shortcuts

and for MPLS you need to enable:

set protocols mpls traffic-engineering bgp-igp

Remember, this deployment is somewhat sloppy and will empty inet.3 and put every thing in inet.0

If you want to have more control over which prefixes should be install in inet.0, use the install and/or active command

HTH

Reza

[Ali Norouzi]

Hi Reza,

Thank you for your reply. I succeed with below configuration on M10.

ali# show protocols
mpls {
    traffic-engineering mpls-forwarding;
    no-propagate-ttl;
    interface all;
}

As you can see I've configured no-propagate-ttl in this router. For some security resons I want to make this router transparent for users.

1- Do you think it's ok to forward all the traffic with MPLS lable towards LSP?

2- Cisco has a feature to enable this feature just for incoming traffic not for traffics that is generated inside the router. You know that is no mpls ip forwarded. Is there same thing in Juniper or not? This feature helps for troubleshooting.

Thank you,

[saurabh joshi]

All traffic originated by RE itself resolves in inet.0 by default.

So if you usemerginng bgp-igp routes that may cause your RE generated traffic to take the LSP but in case you dont do that it will be forwarded based on igp path in inet.0.

Hope this helps..

[Reza Sharifi · Accepted solution]

Hi Ali,

If you want to hide the MPLS network from the customer network use:

"set protocols mpls no-decrement-ttl"

at the ingress LSP only.

If you use "no-propagate-ttl" you have to apply this command to all your routers.

Overall, no-decrement-ttl is easier.

HTH

Reza

[rsimoni · Accepted solution]

I guess you should have posted this on Juniper pages 

cheers,

Riccardo