cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
1130
Views
4
Helpful
7
Replies

Segments not being install at ISIS to OSPF border

Hi all,

I have a network divided into 3 IGP areas:

An OSPF area that is LDP only == An ISIS Area that is SR enabled == An OSPF area that is LDP only

I'm trying to ping from one side to another but SR/LDP interworking is not working at the border between ISIS and OSPF.

Here is my setup:

sr_issue.jpg

All devices can see each others loopbacks. SR is enable in the middle ISIS domain with SRGB range of 16000-20000.
I'm trying to get PE1 - the second to far left device - to ping PE2 (10.1.1.102/32). 

In order to get end to end reachability within the SR domain P2 acts as a mapping server and gives 10.1.1.102 the index of 102. So it should have the segment label 16102. 

And indeed when I trace my way through the network, P2 (which is also in the data path), swaps the incoming label of 24001 (it's local LDP label for 10.1.1.102) with 16102 and forwards it on to R4 in the middle. R4 then carries the segment on and passes the packet to R6 with label 16102 on top. 

But R6 drops the label and when I look at is LFIB it is not installed a local SR label of 16102 for 10.1.1.102 - so it just drops it. And I'm not sure why. It does have an LDP label. But no SR label.

 

 

 

RP/0/RP0/CPU0:P6#show mpls forwarding prefix 10.1.1.102 255.255.255.255
Mon Jul 24 23:37:23.812 UTC
Local  Outgoing    Prefix             Outgoing     Next Hop        Bytes
Label  Label       or ID              Interface                    Switched
------ ----------- ------------------ ------------ --------------- ------------
24016  22          10.1.1.102/32      Gi0/0/0/2    10.6.7.7        0
RP/0/RP0/CPU0:P6#
RP/0/RP0/CPU0:P6#
RP/0/RP0/CPU0:P6#
RP/0/RP0/CPU0:P6#
RP/0/RP0/CPU0:P6#sh isis segment-routing label table
Mon Jul 24 23:37:47.852 UTC

IS-IS CORE IS Label Table
Label         Prefix/Interface
----------    ----------------
16002         10.1.1.2/32
16003         10.1.1.3/32
16004         10.1.1.4/32
16006         Loopback0
16077         10.1.1.77/32
16101         10.1.1.101/32
16602         2001:1ab::2/128
16606         Loopback0
16677         2001:1ab::77/128
16701         2001:1ab::101/128
RP/0/RP0/CPU0:P6#

 

 

 

The next hop for 10.1.1.102 is via OSPF, which is LDP only. But I would expect it to have a local label of 16102 which is would swap out with R7s LDP label as per interworking?

If I strip SR from the middle the pings work fine - end to end. Also, I've shutdown P3 and P5 interfaces so there is only one path through the network (no ECMP or anything).

Any help would be appreciated. Happy to provide full configs/output if needed. 

1 Accepted Solution

Accepted Solutions

Hi @steven.crutchley ,

so in this case, would I take on a kind of seamless MPLS approach?

Yes, this is exactly what I meant. Each domain is completely independent and can run its own IGP and label signalling technique (LDP, RSVP-TE, SR, BGP-LU (inside the domain), etc)

Regards,

Harold Ritter
Sr Technical Leader
CCIE 4168 (R&S, SP)
harold@cisco.com
México móvil: +52 1 55 8312 4915
Cisco México
Paseo de la Reforma 222
Piso 19
Cuauhtémoc, Juárez
Ciudad de México, 06600
México

View solution in original post

7 Replies 7

Harold Ritter
Cisco Employee
Cisco Employee

Hi @steven.crutchley ,

The SR mapping server is normally used to migrate from LDP to SR in a domain with a single IGP. Your scenario is actually invalid. You need to run the same IGP on all routers for the SR to LDP/LDP to SR to work.

If you want to run separate domains with separate IGPs, there are other ways to get it to work, such as running BGP LU between the domains.

Regards,

Harold

Harold Ritter
Sr Technical Leader
CCIE 4168 (R&S, SP)
harold@cisco.com
México móvil: +52 1 55 8312 4915
Cisco México
Paseo de la Reforma 222
Piso 19
Cuauhtémoc, Juárez
Ciudad de México, 06600
México

Interesting... so in this case, would I take on a kind of seamless MPLS approach? Where each ASBR has a BPG-LU session to the other end of this own IGP domain (with next-hop-self turned on for iBGP as appropriate). That way that label for the remote PE can be learned via BGP-LU but within each domain the forwarding is done independently based on whatever label transport is used (LDP or SR) to reach the next hop. 

Hi @steven.crutchley ,

so in this case, would I take on a kind of seamless MPLS approach?

Yes, this is exactly what I meant. Each domain is completely independent and can run its own IGP and label signalling technique (LDP, RSVP-TE, SR, BGP-LU (inside the domain), etc)

Regards,

Harold Ritter
Sr Technical Leader
CCIE 4168 (R&S, SP)
harold@cisco.com
México móvil: +52 1 55 8312 4915
Cisco México
Paseo de la Reforma 222
Piso 19
Cuauhtémoc, Juárez
Ciudad de México, 06600
México

Seamless SR MPLS as @Harold Ritter mentioned is solution. 

Hi @MHM Cisco World ,

Calling this technique Seamless SR MPLS is actually incorrect, as any label signalling is possible. It would be more appropriate to refer to it as Seamless MPLS. Unified MPLS is also commonly used in the Cisco documentation. 

Regards,

Harold Ritter
Sr Technical Leader
CCIE 4168 (R&S, SP)
harold@cisco.com
México móvil: +52 1 55 8312 4915
Cisco México
Paseo de la Reforma 222
Piso 19
Cuauhtémoc, Juárez
Ciudad de México, 06600
México

Hi @MHM Cisco World ,

Seamless SR MPLS implies that SR is used, but Seamless MPLS has been used even before SR was around. In fact, I have seen quite a few deployment of Seamless MPLS with LDP for that matter.

Also, if you are going SR and LDP like the original poster did, this technically can't be referred to as Seamless SR MPLS.

Regards,

Harold Ritter
Sr Technical Leader
CCIE 4168 (R&S, SP)
harold@cisco.com
México móvil: +52 1 55 8312 4915
Cisco México
Paseo de la Reforma 222
Piso 19
Cuauhtémoc, Juárez
Ciudad de México, 06600
México