3750 as a router and a switch. sub-optimal routing problem

[todd.martin-02]

Objective: To limit the traffic on Vlan8. We are seeing traces from K1 or T1 to 192.168.1.4 like this:

1- 10.10.10.5

2- 10.10.20.1

3- 192.168.1.4

then

1- 10.10.10.4

2- 10.10.30.1

3- 192.168.1.4

or

1- 10.10.10.5

2- 10.10.200.1

3- 192.168.1.4

this causes the packets to travers vlan8 once to get routed by P61B then switched back across to reach the next hop 20.1 or vice versa.

How can we avoid this behavior?

Study the attached drawing carefully. Vlan 8 has 4 routers on it and the 3750s have routes to the 4 networks above them. packets are not routed between the 3750's

[bhedlund]

Keep in mind that the router does not always return ICMP traceroute responses out of the interface from which the trace packet was received. The router will make its own independent decision how to send the traceroute response back to you. This can get confusing in L2/L3 multipath switched enviroments like yours. 'debug ip packet' will give you the specific interface traversals.

[todd.martin-02]

Thanks,

I thought that responses DID send the replys back on the interface that it receive the the packet on.

I will provide some debug output on monday. unfortunately I changed the IP's in the drawing to protect the innocent.

From a layer 2 perspective I want to limit the packets traversing vlan8 because that connection is primarily HSRP hello's. It is a SONET connection between bldg's and we want to limit the size of the pipe drasically due to provisioning cost.

[vladrac-ccna]

Hello Todd,

Could set up a debug ip icmp on the router 192.168.1.4 and ping it from both T-1 and K-1, I think this problem would be more clear to everyone.

If not mabybe you could provide the output of "show ip route" on the routes on this issue,at least on the routers/switch 192.168.1.2 , 192.168.1.3 and 10.10.10.4 and 10.10.10.5 and T-1, K-1.

for routes like 192.168.1.4, 10.10.10.2-5 and 20.20.20.2-3

Hope thats not to much info.

Thanks,

Vlad

[bhedlund]

Im not convinced there is a problem here. The IP addresses seen in the traceroute represent the interface the router used to return the trace, not necessarily the interface on which the original trace packet arrived. The only thing I see so far is asymetric routing, which is not abnormal in multipath topologies such as this. Tools such as 'debug ip packet' and 'l2trace' will provide a deeper analysis than traceroute on L2/L3 interface selection in both directions.

Please rate all helpful posts.

Brad

[vladrac-ccna]

Hello Brad,

thanks for the tips on this post, really helpfull for other issues Ive seen.

Vlad

[bhedlund]

Vlad,

L2trace is very cool indeed. Glad I could help.

Regards,

Brad

[bhedlund]

Vlad,

The rating you gave pushed me over 100 points and gave me my first star next to my ID! Thank you!!!

Brad

[todd.martin-02]

I should be able to get some debugs on monday. However the IP's are gonna be different because I changed te IP's in the drawing.