MPLS/default-information originate problem

[d-fernandes]

The following problem happened in my network:

I have two 12406 with IOS Version 12.0(27)S5 (RTR1&RTR2) that make the Internet access from my company through my Service Provider where I have BGP neighbour and I received the Internet full routing table.

All the traffic to Internet goes through the RTR1 and the RTR2 is my backup connection. In RTR1 I have a default-information originate always metric 1 metric-type 1 and in the RTR2 I have a default-information originate always metric 2 metric-type 1.

Also I have in my network two EDGE routers (EDG_RTR1 & EDG_RTR2) and the traffic that goes to Internet through these routers use the default-route advertised by RTR1 and RTR2 respecting the metric.

Also I have two core routers (CORE1 and CORE2) and the traffic that goes to Internet through these routers use IBGP between CORE routers and the RTR routers.

Now the problem:

When I removed from RTR1 configuration the default-information originate always metric 1 metric-type 1 the EDGE routers know the default-route through the RTR2 and the CORE routers continue deliver the traffic to RTR1 but all the traffic to Internet stop in both cases.

The mpls forwarding-table in CORE routers seems to be Ok.

If I removed the tag-switching ip from all interfaces in RTR1 and RTR2 the problem is solved. And if I repeat the process from removing the default-information originate from RTR2 the problem also occur but without the tag-switching ip in interfaces the problem never occur.

Best regards

[mheusinger]

Hi,

in case you remove the default-information command from RTR1, what is the routing table in the core routers showing you for the default route entry?

You could also try to use "mpls ip default-route" to allow label switching of traffic for the default route.

Hope this helps

Martin

[d-fernandes]

Hi Martin,

The core routers have the default route entry through RTR2 but the core router doesn’t use the default-route to deliver the traffic. The core routers have the Internet full routing table.

I have the mpls ip default-route command active.

Regards

[mheusinger]

Ok,

the "mpls ip default-route" command allows label switching for the default route. So the internet routing table on the core routers will not be used (no IP lookup, just label lookup), if you have MPLS to the edge routers.

Can you check the LFIB and CEF entries for the default route in the core routers under the various conditions?

By the way, removing the command from all routers will revert to IP forwarding when the default route is matching. In case there are issues with label switching for the default route you could safely remove the command.

Hope this helps

Martin

[d-fernandes]

Martin,

I saw the LFIB and CEF entries for the default route in the core routers under the various conditions and it seemed to be Ok. Due to loss of service that occurred I not be able to make more exhaustive troubleshooting.

But I will replicate in laboratory my network with same router platforms and same IOS to make a more exhaustive troubleshooting.

Meanwhile I found a bug very similar to my problem:

-----------------------------------------------------------------------------

CSCsa64782

Symptoms: When an ingress ISE line card is used with a default route that iBGP

learns over a MPLS core, the following two symptoms may occur:

- The output of the show controllers tofab alpha mip stat | i

MTU command may show traffic drops.

- Traffic is incorrectly sent as "unlabeled" over the MPLS core.

Conditions: These symptoms are observed on a Cisco 12000 series when the

traffic path follows a recursive default route and when recursive load sharing

occurs.

Workaround: Prevent outbound load sharing to the default route by changing the

IGP metrics.

-----------------------------------------------------------------------------

However I don’t have load sharing .

Regards

[olorunloba]

I'm not entirely clear of the scenario, but this is what I can deduce. When working with the default-information configured the edge routers learn the default route via OSPF, and you hve labelling for the default route as well. I assume also that you have an unbroken lsp from the edge routers to the internet gateways, else, route lookups in the core will distribute the internet traffic amongst the two gateways (depending on the internet BGP table).

I will like to assume that after removing the default-information config, that the outgoing label on the edge routers (for the default route) is no longer the same as in the incoming label on the attached core router. This could be a bug. I'm not sure of what the scenario will be like if the core routers are learing the default route via iBGP (but the problems seems possible as well).

To properly troubleshoot, I will advice that you analyse the output of sh ip cef 0.0.0.0 0.0.0.0 details, before and after removing the default-information on all the routers. Also check the sh ip cef details for the BGP router-ids of the gateway routers on all the routers. There should be a disparity somewhere that is leading to the packets being dropped. You might want to give us this outputs so we can learn also.

[olorunloba]

Also try clearing the ip route table, to see if it solves the problem.