MPLS Traceroute - P Router Asterisk (*)

[Andrew Xue Zheng Lay]

Hi,

I need some help in understanding MPLS Traceroute.

My P Routers are displayed as Asterisk (*), and I'm not sure why.

And before someone comment about this... I do not have "no mpls ip propogation-ttl"

SPOKE_B_USER#traceroute 172.16.1.11 probe 1
Type escape sequence to abort.
Tracing the route to 172.16.1.11
VRF info: (vrf in name/id, vrf out name/id)
  1 172.16.2.17 1 msec (Spoke B's Core SW)
  2 172.16.2.1 0 msec (Spoke B's CE)
  3 123.45.67.33 9 msec (ISP's PE2)
  4  * (ISP's P2)
  5  * (ISP's P1)
  6 124.45.67.21 [MPLS: Label 16 Exp 0] 9 msec (ISP's PE1)
  7 124.45.67.22 9 msec (Hub's CE)
  8 123.45.67.21 10 msec (ISP's PE1)
  9 123.45.67.5 [MPLS: Labels 19/30 Exp 0] 15 msec (ISP's P1)
 10 123.45.67.29 [MPLS: Label 30 Exp 0] 10 msec (ISP's PE3)
 11 123.45.67.30 18 msec (Spoke A's CE)
 12 172.16.1.2 15 msec (Spoke A's Core SW)
 13 172.16.1.11 18 msec (Spoke A's USER)

[Peter Paluch]

Hi Andrew,

Would you mind trying the traceroute without the probe 1 argument? It is possible that the ISP's P-routers are throttling down ICMP messages, and your messages got lost because of that.

If using 3 or more probes per hop does not help then it is possible that the ISP's P-routers are configured not to send ICMP unreachable messages. That would explain why they don't show up in your traceroute. Also, if they're not Cisco routers, it is remotely possible that instead of sending the ICMP unreachable message along the original Label Switched Path, they try to send the ICMP unreachable mesage back to the sender directly which may not be possible because of several reasons.

Would any of this explain your observations?

Best regards,
Peter

[Andrew Xue Zheng Lay]

Hi Peter,

I've found the cause of the issue actually. But... I do not understand why it is happening that way.

This is supposed to be like a Hub and Spoke Topology by the way. Spoke to Spoke traffic must pass through Hub.

On PE1, I've added

PE1(config)#ip vrf SpokeA
PE1(config-vrf)#route-target import 65002:65002

And the P Router hops started to appear.

SPOKE_B_USER#traceroute 172.16.1.11 probe 1   
Type escape sequence to abort.
Tracing the route to 172.16.1.11
VRF info: (vrf in name/id, vrf out name/id)
  1 172.16.2.17 0 msec
  2 172.16.2.1 1 msec
  3 123.45.67.33 9 msec
  4 125.45.67.22 [MPLS: Labels 20/16 Exp 0] 10 msec
  5 125.45.67.22 [MPLS: Labels 21/16 Exp 0] 9 msec
  6 124.45.67.21 [MPLS: Label 16 Exp 0] 10 msec
  7 124.45.67.22 10 msec
  8 123.45.67.21 9 msec
  9 123.45.67.5 [MPLS: Labels 19/30 Exp 0] 18 msec
 10 123.45.67.29 [MPLS: Label 30 Exp 0] 7 msec
 11 123.45.67.30 18 msec
 12 172.16.1.2 18 msec
 13 172.16.1.11 18 msec

Why is that affecting it?

[Andrew Xue Zheng Lay]

Found the actual cause of the problem.

CE-H is running NVI "ip nat enable".

Swapped it with regular NAT "ip nat inside" and it resolves the issue.

I'm not sure how or why it will affect the MPLS traceroute though.

[Peter Paluch]

Hi Andrew,

Expletives are not necessary - we understand your frustration just as well without them :)

Anyway, can you post the configuration of the CE-H with the NVI-style of NAT that didn't work? It's difficult to comment on why it didn't work properly without seeing the details of the configuration.

Thanks!

Best regards,
Peter

EDIT: It might actually be useful to post the configuration of all devices on the path. The impact of addition of the route-target import on your PE1 suggests that a possible reachability issue exists as well. Also, after you edited your posts, the diagram of your network was removed - can you submit it again?

[Andrew Xue Zheng Lay]

Nah... I think it's fine now.

I have no problems sticking to the traditional NAT.

I didn't have good experience with NVI since the past anyway.

Did a LAN-Internet NVI with Static Mapping on a 2611XM a couple of years back, turns out that performance dropped more than 50%.

But who knows... maybe I'm not using it right.

But thanks anyway.

[Peter Paluch]

Andrew,

Yeah, the NVI NAT does have peculiarities, and is poorly documented - the details of exactly "when" and "how" are not well covered.

Anyway, I'm glad you got it running.

Best regards,
Peter