OSPF with spanning tree

[carl-voicenet]

Hi there Merry Xmas..

We have two 3750 core switches and one mpls switch.

Core Switch B connects to switch A via port channel. The Mpls switch connects directly to switch A.

Switch A is hsrp active for all vlans.

We would like to connect the MPLS switch to Core switch B as well and use that for redundancy.

Now i understand also that spanning tree will be an issue with 2 connections from the mpls switch.

I've setup ospf on all devices and core switch A is the spanning tree root.

The issue though is that traffic across the MPLS network seems to work fine to start with then after about 20mins it stops routing across. I then have to unplug the MPLS connection to switch B and then reset the connection to switch A to fix.

Any thoughts?

Is this best method to obtain redundancy for this?

OSPF on just core switch A works fine but once adding core switch B into ospf we start to see issues.

Basic Diagram attached..

[amaragra]

What as such do you mean by 'stops routing across'... can you please explain (what did you do to test this).

Once the connection is done between switchB and the mpls switch you will have to find out which port is in spanning tree blocked state as this will tell you the traffic path from internal network to the mpls network.

if after making a connection between switchB and MPLS switch the network goes down again, you will have to do the ping test hop by hop and using routing table you can find out what is the issue

[carl-voicenet]

A ping test was done from a PC connected to core switch A across the MPLS network. After about 20mins or so this would drop and i would need to reset the connections to the MPLS switch. No idea why this happens. But it did it twice.

[JoeKeegan3]

Hi Carl,

It's hard to follow exactly what the problem might be with your description, but I did have a suggestion for the design.

It's not clear if the MPLS switch is a L3 switch, but if that is the case then it would be best to make the connections between the the MPLS switch and your core switch routed ports and not switch ports and simply route between the three switches. This will remove spanning tree from the equation.

if the MPLS is not a L3 switch then I would look at having the connections from the core to the MPLS included in a VLAN, but have the connection between the two switches as a routed connection. This would remove any looped topology. In this configuration you would have to make sure to cost the VLAN connections higher to make sure routing between the cores went over the directly connection instead of via the MPLS switch.

I've attached a simple diagram to show what I am talking about.

If you want to figure what issue you are having then it would be good to see the relevent configurations of the devices and maybe the output of some show commands like show spanning tree and show ip osf.

All the best,

Joe

[carl-voicenet]

Thanks for the reply.

The mpls switch is currently sitting in a vlan with the core switches but we are able to make these routed ports. Whats the advantage of this?

Does it matter that both routed ip addresses on the MPLS switch will be in the same subnet range?

Whats the best method to set this up for redundancy once routed ports setup? Do i need to set priority or cost on ospf?

[JoeKeegan3 · Accepted solution]

The mpls switch is currently sitting in a vlan with the core switches 
but we are able to make these routed ports. Whats the advantage of this?

There are a few advantages, but most of it is related to minimizing spanning-tree and removing looped L2 topologies. When you have a design that includes a looped L2 topology then spanning tree will block some links, this will reduce the efficency of the network (since the most direct path maybe blocked) and you need to make sure that L2 and L3 topologies are "in sync". For example if you are not careful with your root brigdge configuration then you could end up with one core switch needing to go through the other core switch to reach the MPLS switch, instead of both using thier path conencted to the MPLS switch.

Another factor is that each L2 domain represents a failure domain, where if you have L2 issues (broadcast storm, STP failure, etc) that can cause the entire VLAN to fail, which could bring down your connection to the MPLS for both core switches.

There are some other benifits such as faster failure recovery, but In general spanning-tree and L2 looped topologies should be avoided especially in your core.

Does it matter that both routed ip addresses on the MPLS switch will be in the same subnet range?

I'm not 100% sure what you mean by this, but if you have all the links as routed ports, you should use a unique /30 or /31 for each link.

Whats the best method to set this up for redundancy once routed ports setup? Do i need to set priority or cost on ospf?

As long as they are all routed ports then you don't need to make any changes on the cost or priority, OSPF will figure out the opimal path to use. This assumes that everything is using the same links, i.e. they are all GigE, etc.

[carl-voicenet]

Yup thats sounds good. I'll go with that..

Many thanks..

[JoeKeegan3]

Best of luck and have a good new years!