Dear Jon thanks a lot for the prompt reply, thats exactly what i want to know that what are the cases in which router DOESNT assign the labels to prefixes and these prefixes are seen as UNTAGGED on the neigboring router !

Kindly help me out

Well the action will be untagged if the outbound interface on the router is not mpls enabled so mpls tags would make no sense.

You might also want to look at a thread i posted and got answered on the MPLS forum a while back -

http://forum.cisco.com/eforum/servlet/NetProf?page=netprof&forum=Service%20Providers&topic=MPLS&topicID=.ee8558c&CommCmd=MB%3Fcmd%3Dpass_through%26location%3Doutline%40%5E1%40%40.1dd9f838/7#selected_message

Hope i've answered your question.

Jon

Dear Jon kindly look at my following scenario

R1-----R2-----R3

|

|

|(10.0.0.0/8)

R4

1) ospf is enabled on all interfaces

2) mpls is enabled on R1, R2, R3.

3) mpls is NOT enabled between R1 and R4.

Now Jon like you said, at R2 i should see prefix 10.0.0.0/8 UNTAGGED right ? since R1 is not running mpls on this link, thats exactly what i thought, but when i try it on router, it showed me POP tag on R2 !! pls explain why, coz i have heard ppl saying that if mpls is not enabled on a link then its advertised with no label and hence untagged but i am seeing pop tag, pls look into this,

i am newbie to mpls, pls help me out

Dear Jon kindly look at my following scenario

R1-----R2-----R3

|

|

|(10.0.0.0/8)

R4

1) ospf is enabled on all interfaces

2) mpls is enabled on R1, R2, R3.

3) mpls is NOT enabled between R1 and R4.

Now Jon like you said, at R2 i should see prefix 10.0.0.0/8 UNTAGGED right ? since R1 is not running mpls on this link, thats exactly what i thought, but when i try it on router, it showed me POP tag on R2 !! pls explain why, coz i have heard ppl saying that if mpls is not enabled on a link then its advertised with no label and hence untagged but i am seeing pop tag, pls look into this,

i am newbie to mpls, pls help me out

Hello Ovais,

when you enable MPLS on a router:

the router will start to send out LDP discovery messages out configured interfaces to find neighbors.

AND

for each prefix present in the CEF table and known by an IGP, or connected and so on will assign a local label.

So in your case R1 will build an LDP neighborship to R2 and not to R4 because MPLS is not running on the link between R1 and R4.

But R1 to R2 sends out its label bindings including an MPLS implicit null for net 10.0.0./8.

From the point of view or R2 the same prefix is reached via a POP tag action.

b) untagged

The untagged action can be seen when you use label filtering:

the default action is to assign a label to every known prefix but this is not necessary in a service provider network using IGP + iBGP: in this case the label space can be saved by configuring all the nodes to assign labels only to specific IP addresses blocks that are those of the loopbacks used for iBGP.

with the following commands

mpls advertise-tags for PREFIX-LDP-out

ip access-list standard PREFIX-LDP-out

permit 10.80.0.0 0.0.255.255

in our network we instruct all routers to create labels only for prefixes inside 10.80.0.0/16 that are for loopbacks.

in this way for example all routers agree on action Untagged for prefix 10.104.160.0/24.

Hope to help

Giuseppe

Dear Guistar, i am near to clear my confusion, if i am practicing basic mpls without any vpn and advance stuff, then can you give me some guidelines as to which prefixes a router will not define label for ? is it correct that prefixes outside mpls domain (except the directly connected interfaces) will not be assigned any labels ?

Am i correct ?

Hello Ovais,

the router will try to bind a label to every known prefix in its routing table (CEF table is the right term/concept for MPLS depends on it).

see the following example

RT-RM-TLD066-RG-MAN-E-2#sh mpls forw 172.24.154.0

Local Outgoing Prefix Bytes tag Outgoing Next Hop

tag tag or VC or Tunnel Id switched interface

1214 Untagged 172.24.154.0/23 0 Vl617 10.82.8.108

Untagged 172.24.154.0/23 0 Vl617 10.82.8.109

RT-RM-TLD066-RG-MAN-E-2#

RT-RM-TLD066-RG-MAN-E-2#sh ip route 172.24.154.0

Routing entry for 172.24.154.0/23

Known via "eigrp 100", distance 90, metric 63488, type internal

Redistributing via isis, eigrp 100

only because there is a filter as showed in my previous post a label is not assigned for this subnet known via EIGRP on a link out of the MPLS cloud:

RT-RM-TLD066-RG-MAN-E-2#sh mpls int vlan617

Interface IP Tunnel Operational

Vlan617 No No No

RT-RM-TLD066-RG-MAN-E-2#

So one thing is to be part of the MPLS cloud and to build LDP adjacencies, another aspect is the signalling (of what IP networks the router will advertise label binding).

This happens also for static routes

:

RT-RM-TLD066-RG-MAN-E-2#sh ip route 172.18.164.0

Routing entry for 172.18.164.0/24

Known via "static", distance 1, metric 0

RT-RM-TLD066-RG-MAN-E-2#sh mpls forw 172.18.164.0

Local Outgoing Prefix Bytes tag Outgoing Next Hop

tag tag or VC or Tunnel Id switched interface

37 Untagged 172.18.164.0/24 0 Vl900 10.98.164.18

RT-RM-TLD066-RG-MAN-E-2#

RT-RM-TLD066-RG-MAN-E-2#

RT-RM-TLD066-RG-MAN-E-2#sh mpls int vlan900

Interface IP Tunnel Operational

Vlan900 No No No

RT-RM-TLD066-RG-MAN-E-2#

The objective is to implement a separation of signalling plane and forwarding plane: from the point of view of the MPLS cloud all IP prefixes out the cloud are leafs originated by a LER = Label Edge Router without looking at the route origin on the LER node.

So in your case if there is a prefix behind R4 and it is advertised to R1 you can expect R1 to advertise it with an implicit null label to R2.

Hope to help

Giuseppe

Dear Guistar, so am i correct in this that, router will advertise imp null in 3 cases

1) for all directly connected networks whether mpls is enabled on them or not

2) for summary routes (taken from mpls and vpn arch )

3) for all networks learned from non-mpls neighbors

Am i correct in this ?

Dear Msobier, thats exactly what i am confused about,

When router will not assign a label to prefix ?

Hi,

As this thread is more active I'll repost my answer from the MPLS forum, which I found first.

Hi,

It is inevitable for all MPLS environments that there will be pure IP networks, which get MPLS labels assigned. It can be explained in different ways. Answers to your question:

A1) MPLS is a transport technology transparent to the IP end systems (like PCs). Thus an edge MPLS router will have to assign labels to "pure IP networks". Otherwise no MPLS labels could be assigned to those networks which traffic should be transported.

A2) MPLS label allocation - by default - follows a simple rule: allocate a label to every IP route except if it is learned by BGP. Locally assigned labels are sent to all LDP neighbors. A MPLS router does neither know, if a neighbor will ever use those labels nor, if it is essentials to have a label. Assume as an example you have configured MPLS L3VPN with your R1 routers 14.0.0.0/8 address as next hop. Then it would be crucial not to use "unlabeled/untagged", as this would break connectivity. LDP cannot "know" BGP next hops used somewhere in your network. Therefore the default label allocation scheme is as described and observed.

Note: LDP can be configured with selective label allocation, as the network administrator knows, what LDP can not. In your example you could restrict LDP to only allocate/announce labels for loopback addresses. Then the behaviour you expected from the beginning could be observed. As depicted in A2) misconfigured selective label allocation can break connectivity in MPLS networks.

Hope this helps! Please use the rating system.

Regards,

Martin