Routing Mystery - Help understanding

[Nicholas Sansone]

Good afternoon,

We have two 3850 switch stacks:

Switch Stack A, Switch Stack B, and our 2 core switches

There is a connection between Switch Stack A and Switch Stack B

-----Stack A Config----- > Connects to TenGigabitEthernet1/1/3 on Stack B

interface TenGigabitEthernet1/1/4

description "Link to Switch Stack B"
no switchport
ip address 10.5.152.21 255.255.255.0
end

-----Stack B Config-----

interface TenGigabitEthernet1/1/3
description "Library"
switchport access vlan 152
switchport mode access
end

We have EIGRP configured, it is just set for network 10.0.0.0/8, there are no additional configurations, here is a list of some of the routes

D 10.20.40.0/22
[90/3072] via 10.5.152.20, 7w0d, TenGigabitEthernet1/1/4
D 10.20.45.0/28
[90/3072] via 10.5.152.20, 7w0d, TenGigabitEthernet1/1/4
D 10.20.56.0/22
[90/3072] via 10.5.152.20, 7w0d, TenGigabitEthernet1/1/4
D 10.20.60.0/24
[90/3072] via 10.5.152.20, 7w0d, TenGigabitEthernet1/1/4
D 10.20.64.0/24
[90/3072] via 10.5.152.20, 7w0d, TenGigabitEthernet1/1/4
D 10.20.65.0/24
[90/3072] via 10.5.152.20, 7w0d, TenGigabitEthernet1/1/4
D 10.20.92.0/24

The route they are hopping to is 10.5.152.20, which is an IP address on the Core Switch stack. I don't understand how this is working, the ports seem to be misconfigured.

Stack A > Stack B > Core for routing

But the link between Stack A and Core doesn't seem to be present somehow, or via some loophole the IP address interface on Stack A combined with EIGRP being on is allowing it to see the core?

Also, EIGRP is NOT turned on for Stack B that is connected to the core. It is marked as no switchport, so that should stop it from creating a bridge between both stacks and the core.

Any help understanding this would be greatly appreciated. 

[Richard Burts]

We do not have enough information to be able to understand what is your issue or to give advice about how to solve it. You tell us about two stacks and two core switches. You give us interface configuration of a single interface in each stack but nothing about the core switches. We need more information if we are to be able to help you.

There does seem to be some mismatch between the configuration of the interfaces that you posted. stack A is no switchport while the connected interface on stack B is switchport. I am not sure why it is set up this way but it does not prevent communication through that connection. It is apparent that EIGRP messages from stack A are getting through the switch in stack B and on to which ever of the core switches is 10.5.152.20.

HTH

Rick

[Nicholas Sansone]

These are the only two interfaces that are connected between the stack and there are no other ports connected that go between the stacks, so I didn't think I would need to provide anymore configuration information. 

Stack B (Routing port) >>>>>> Stack A (Access port) >>>>>> Core (Contains the last hop in all of the routes that are listed in the routing table on Stack B)

-----From Stack A------

!

router eigrp 1
network 10.0.0.0
eigrp stub connected summary
!

All the traffic knows to go out this routed port on Stack A, but it isn't directly connected to the Core to learn about it, so it must be learning about it via EIGRP. My confusion is how the EIGRP packets are getting through the access port to the core

In all of my labs/training, it was always setting up EIGRP between two routers and setting an IP interface on both sides (Stack A interface gets an IP, Stack B interface gets an IP, etc)

[Alex Pfeil]

Please draw out the design of the four devices.

Then run,

show ip interfaces brief

show ip eigrp neighbors

show ip eigrp interfaces

show ip eigrp topology

[Richard Burts · Accepted solution]

You still are not providing information to allow us to understand what is going on here. From the little bit that we know I believe that this is what is happening:

- stackA interface 1/1/4 is configured as a routed port (no switchport) with address 10.5.152.21

- stackA is running EIGRP on that interface

- so stackA is sending EIGRP packets out interface 1/1/4

- stackB interface 1/1/3 is a switchport in vlan 152

- stackB receives EIGRP packets on 1/1/3

- stackB forwards EIGRP packets to other interfaces in vlan 152 (and if there are trunks that carry vlan 152 then the EIGRP is carried on them)

- one of the core switches seems to be connected to stackB and receives vlan 152 packets on an interface with IP 10.5.152.20

- core switch is running EIGRP on that interface

So what can you tell us about vlan 152 on stackB? Does it have other interfaces in vlan 152? is it carried in a trunk?

What can you tell us about connections between stackB and the core? Is it access ports, is it trunk?

HTH

Rick

[Nicholas Sansone]

What you have typed out is correct.

Routed port on B >>>> Switched port on A / VLAN 152 (This is part of a trunk that goes to the core) >>>> Core
So in this sense then, Switch Stack B is just using Switch Stack A as a switch to essentially just carry EIGRP packets to the core. Because the core can see these packets, it is passing EIGRP packets back and forth through this switch.

This all makes sense now. The better design we are looking to implement is to just bypass Stack B and go straight from Stack A to the Core, as there is no real reason to go through Stack B.

Thanks for breaking it down, that makes sense now, its just using that VLAN as a trunk to carry the data to the core.

[Richard Burts]

I am glad that my explanation has been helpful. With two stacks and two cores it would seem logical that you might want some redundancy so that stackA could get to cores directly or could use the path through stackB. Or is there some reason why you would want stackA to have only a single path to the cores? There is also perhaps some question about what role (if any) you intend for stackB in the routing logic for your network.

HTH

Rick

[Nicholas Sansone]

OK, I think i figured out part of the solution, the stack is configured as an EIGRP Stub, so this means that it is funneling all of the routes to the distribution EIGRP router, which is the core. I'm still confused as to why it can just pass through the other switch stack on a port that is tagged as an access port to a VLAN that isn't on Switch Stack A.

[kumowoon1025]

What does show ip arp look like on A? I think maybe Switch B has ip proxy arp?