Cisco OSPF - ECMP for redistributed routes

[Sebastian Narp]

Hallo,

i have two Cisco Nexus 5500 - both are connected via OSPF and both have an OSPF Partner which knows a route to one destination:

nexus01# show ip route 135.138.100.10 vrf FE
IP Route Table for VRF "FE"
'*' denotes best ucast next-hop
'**' denotes best mcast next-hop
'[x/y]' denotes [preference/metric]
'%<string>' in via output denotes VRF <string>

135.138.100.10/32, ubest/mbest: 1/0
    *via 10.5.8.18, Vlan58, [110/20], 05:00:41, ospf-FE, type-2

nexus02# show ip route 135.138.100.10 vrf FE
IP Route Table for VRF "FE"
'*' denotes best ucast next-hop
'**' denotes best mcast next-hop
'[x/y]' denotes [preference/metric]
'%<string>' in via output denotes VRF <string>

135.138.100.10/32, ubest/mbest: 2/0
    *via 10.5.6.66, Vlan56, [110/20], 00:00:03, ospf-FE, type-2

Every of them has type-2 routes, the costs are both the same. So why nexus01 has only one entry in his routing table?

He can reach 135.138.100.10 over two routers with the same cost...

Thanks for your help!

Best Regards, Sebastian

[Rolf Fischer]

Hallo Sebastian,

unfortunately I'm not familiar with NX-OS.

We'll need to have a look at the Type-5 LSAs and the distances to the ASBRs.

In IOS, this would be the useful commands:

• show ip ospf database external 135.138.100.10
• show ip ospf border-routers

Could you try to find the corresponding commands on your Nexus and share the output?

HTH
Rolf

[Sebastian Narp]

Hallo Rolf,

thanks for your help, the command is the same but you need the correct vrf:

nexus02# show ip ospf database external 135.138.100.10 vrf FE
        OSPF Router with ID (10.5.2.3) (Process ID FE VRF FE)

                Type-5 AS External Link States

Link ID         ADV Router      Age        Seq#       Checksum Tag
135.138.100.10  10.5.3.18       739        0x8000002f 0x4542    0
135.138.100.10  10.5.7.9        163        0x800000c0 0x3cbe    0

We see both links, the ADV Router IDs are our two OSPF neighbors on the switches. Over both we need an ECMP.

nexus02# show ip ospf border-routers vrf FE
OSPF Process ID FE VRF FE, Internal Routing Table
Codes: i - Intra-area route, I - Inter-area route

i 10.5.3.18 [100], ASBR, Area 0.0.0.0, SPF 395
     via 10.5.3.18, Vlan58
i 10.5.7.9 [200], ASBR, Area 0.0.0.0, SPF 395
     via 10.5.0.66, Vlan55
     via 10.5.0.74, Vlan56

Those two have differnt costs. But because of the Routetypes this should not be an Problem, correct?

Thanks for your help!

Sebastian

[Rolf Fischer · Accepted solution]

Sebastian,

Those two have differnt costs. But because of the Routetypes this should not be an Problem, correct?

This is actually a very common misinterpretation of E2 routes, so don't worry ;)

Yes, they all show the same (default) metric in the routing table but this is not the criteria for ECLB.

Have a look at the E1 vs. E2 External Routes section of the OSPF Design Guide[*]:

If the external routes are both type 2 routes and the external costs to the destination network are equal, then the path with the lowest cost to the ASBR is selected as the best path.

HTH
Rolf

[*]: Or section 16.4 (6) of RFC 2328

P.S.: More CSC discussions on the E2/E1 Topic:

https://supportforums.cisco.com/discussion/11611716/what-point-external-e2-ospf-types

https://supportforums.cisco.com/discussion/12253961/ospf-selection

[Sebastian Narp]

ahh okay, thanks a lot.

So, if I understand you correct, ECMP would only work in even the ASBR is the same?

How can I ECMP if I have to different ASBRs?

I think the problem is not to find information about OSPF, i think the problem is to find the information when ECMP works or how it works.  How do you find the information about ECMP?

Thanks again, and thanks a lot more for your patience!

Sebastian

[Rolf Fischer]

You're welcome.

I have attached a screenshot of one of my mindmaps which shows the decision process for external routes as defined in section 16.4 of RFC 2328. I hope it's easier to understand but it is in my native language German, so let me know if I'm wrong and you don't understand German ;)

Not sure if that answers your question or if your're asking more general about how a Cisco device performs ECMP. In that case this document may be a good start: Troubleshooting Load Balancing Over Parallel Links Using CEF

Rolf

[Sebastian Narp]

Hey Rolf,

thanks a lot, the mindmap is great. I'm a native german speaker so its perfect :)

So in our case - different ASBR and different internal costs - we have no possibility to make ECMP. Is this correct?

Vielen Dank und einen schönen Arbeitstag wünscht dir

Sebastian

[Rolf Fischer]

So in our case - different ASBR and different internal costs - we have no possibility to make ECMP. Is this correct?

With external type 2 LSAs, yes, absolutely. The LSA metrics need to be equal and the (internal) costs to the FAs/ASBRs as well in order to load balance.

Type 1 external LSAs require, in contrast, just equal total costs - the respective components can be different. By the way: As far as I can say, most people prefer Type 1.

Danke, ebenso :)

Rolf

[Sebastian Narp]

Okay, thanks a lot for your help! Now I understand my mistake and I think we cant solve our problem that way. We have to think about that and maybe redesign some thinks...

Best regards,

Sebastian