How does GLBP and Spanning-Tree affect traffic flow when running together?

[dimzaaaaa]

Guess the web ate half the diagram but well...Five Levels from Top to Bottom:

1-Firewall

2-Layer 2Switch--Tie Between 2 Cores and Firewall

3-Two Main Cores--Pruned Layer 2 Port-Channeled Trunks carry Layer 3 Traffic(Platforms won't support sub-interfaces for integrated vrf configuration)

4-Layer 3 Distribution Points(layer 3 devices separating key sections)--again pruned Layer 2 Port Channels

5-Access Switches-Layer 2 Trunks to End users.

Specific question is unique to the links between Levels 2 and 3(Layer2 Switch To  Two Cores Connection)

Now supposing that spanning-tree protocol is running between the three devices in a triangle(Layer2 Switch the Vertex), and that Cores A and B are running GLBP protocol to service some subnets(approximately vlans) which,(due to transitional inconvenience) have to be routed from the Core instead of the Sectional Distribution points(Level4 Switches) devices; What device needs to be set up as the SPANNING-TREE ROOT(r-pvst), thus

letting GLBP to properly load balance between the two Core Devices?

Keep in Mind Both Cores are connected to each other, and then to the layer 2 switch.

I need to get independent opinions so as to reduce the chance of misconfiguration mishaps, since whatever I set up is gonna be pretty much it, and I will be expected to clean up any projected messes depending on what I opt for.

As I post the question I will take the liberty to revisit the GLBP white papers on Cisco and perhaps something about RSTP to go along. Any input is much appreciated.

[Jon Marshall]

dimzaaaaa wrote:
Specific question is unique to the links between Levels 2 and 3(Layer2 Switch To  Two Cores Connection)
Now supposing that spanning-tree protocol is running between the three devices in a triangle(Layer2 Switch the Vertex), and that Cores A and B are running GLBP protocol to service some subnets(approximately vlans) which,(due to transitional inconvenience) have to be routed from the Core instead of the Sectional Distribution points(Level4 Switches) devices; What device needs to be set up as the SPANNING-TREE ROOT(r-pvst), thus
 letting GLBP to properly load balance between the two Core Devices?
Keep in Mind Both Cores are connected to each other, and then to the layer 2 switch.

The vlans that have to be routed on the core, are they for devices directly connected to the core or are they for devices on others switches (not shown in your diagram) that are dual connected to each core switch ?

If the devices are directly connected into the core switches are they dual honed or singly connected to one or other of the core switches ?

Basically STP will first create a loop free topology and then GLBP will run on top of that. As to how to set that up really depends on the questions above.

Jon

[dimzaaaaa]

Jon,

The affected Vlans needing Layer 2 connections to the Core are an off-shoot of level 5 switches(access-switches); connected to their area Distro(level 4) switches(single link); the Distro switches are then dually connected to the Cores(level3); the Cores then run GLBP between each other thus effectively route for those vlans. The specific dilemma is that, in Spanning Tree's effort to create a Loop Free Topology, does it then follow that GLBP is going to be able to do it's load balancing for the traffic taking into account the Spanning Tree Blocked ports??

[Jon Marshall]

dimzaaaaa wrote:
Jon,
The affected Vlans needing Layer 2 connections to the Core are an off-shoot of level 5 switches(access-switches); connected to their area Distro(level 4) switches(single link); the Distro switches are then dually connected to the Cores(level3); the Cores then run GLBP between each other thus effectively route for those vlans. The specific dilemma is that, in Spanning Tree's effort to create a Loop Free Topology, does it then follow that GLBP is going to be able to do it's load balancing for the traffic taking into account the Spanning Tree Blocked ports??

Yes GLBP will still be able to load-balance but it does mean that the interconnect between our cores will see a lot of traffic assuming one of the uplinks from the distro switches is being blocked. You could load-balance vlans from the distro to the core by alternating the STP root for odd and even vlans between the core switches but if you do this then you may as well use HSRP and just match up the HSRP active with the STP root for any particular vlan.

So you need to make sure that the core switches interconnect is scaled enough to take the additional traffic if using GLBP and it will use a less than optimal switching path for some traffic. The only way you could utilise both uplinks would be to make the interconnect between your 2 core switches a L3 link but this is not a trivial thing to do without planning and may not suit your topology.

Jon