Wired Intelligent Edge

If i have a VSF enabled 5400R core, and I want to connect stacked distribution switches to them via a trunk, there is only one logical forwarding path.

What is the recommendation STP wise? Best practice?

Enable it anyway on both core and distribution?

Make the VSF core the STP root?

If running MSTP, would it be best to just dump all VLANs in instance 0? Or create a new instance 1 for example?

Hi Pepe,

I also have the same question, see this post:

http://community.arubanetworks.com/t5/Campus-Switching-and-Routing/VSF-and-Link-Aggregation-doubt/m-p/363154#M2655

Although the ACSA Official Certification Study Guide says the following:

Hello!

1. Normally, the root of the STP topology is the CORE switch of the network because it has more processing resources. It is interesting to set the primary and secondary root.
2º When you use VSF, you will have a unique logical switch in the network. And we can imagine the same case for link-aggregation (trunk). Logic you are interconnecting a 2GB cable into a single switch (2 cables 1GiG).
3º For the STP topology, the aggregation connection is seen as a single logical link.
4º The only premise for this activity is that you enable the STP on both switches before interconnecting the aggregation cables.
5º If you are unsure, perform the aggregation / trunk configuration on both switches, enable the STP, and interconnect only 1 cable. After, see the result of the "show spanning-tree" command on both switches. You will see that the STP will detect the logical interface (trk) and not the physical interface.

This scenario will work perfectly !!!

John,

Very good explanation!

Regards,

Julián

Thanks for the fast reply.

The situation is as follows. I have now 2 5412zl cores,which need to be replaced by the R version.

I chose stacked distribution switches as to have one logical link to the new VSF core (to replace current distribution switches which are 2 in each building).

Currently we have just "spanning-tree" enabled on all switches (including access layer). Which ties in to my other post regarding what is the default spanning tree operation out of the box.

You are right that if i don't have full control over the rooms/cabling where the core switches reside, it's best to turn on spanning tree on the new Cores.

But on all of our other sites we are running MSTP with the idea of load balacing on L2 via 2 instances. This is our standard.

So first of all i would like to know:

- is there a documented statement somewhere which says "Always turn on STP on your VSF core in case **** happens" (or don't turn it on because it's not needed if you have only one logical link)

- if i don't have a need for load balancing, is there even a need for turning on MSTP explicitly, like the standard defines on my other sites? If the default is MSTP with instance 0, or RSTP, i have fast convergence anyways, so then i don't have to configure ALL my switches (including access layer) in instance 1 for example. I could just leave the spanning-tree command on all switches enabled which would have my entire network/VLANs in instance 0.

- Is there a HARD/Documented reason to have your VLANs placed in another instance than 0? Is there a danger so to speak?

For now my thinking is: Attach my new VSF core to the old Cores. Move SVIs from old to new with the "disable/enable layer3 vlan" command. And make the single VSF core have the command "spanning-tree" and "spanning-tree prio 0". Just leaving all of my VLANs in instance 0. Same on distribution layer.

I think it would be a bit redundant to map all my VLANs in instance 1 for example, if they are all located in instance 0 anyway and i don't need L2 loadbalancing.

- is there a documented statement somewhere which says "Always turn on STP on your VSF core in case **** happens" (or do not turn it on because it is not needed if you have only one logical link)

"From the moment you have redundant paths, you should think about working as STP. There is no right answer to that, it all depends on your network, your knowledge and your mappings."

- if i do not have a need for load balancing, is there even need to turn on MSTP explicitly, how does the standard define on my other sites? If the default is MSTP with instance 0, or RSTP, i have fast convergence anyways, so then i do not have to configure ALL my switches (including access layer) in instance 1 for example. I could just leave the spanning-tree command on all switches enabled which would have my entire network / VLANs in instance 0.

"Right."

- Is there a HARD / documented reason to have your VLANs placed in another instance than 0? Is there a danger to speak?

"If you want to do Vlans balancing with MSTP, you need to create other instances. If you leave all Vlans in instance 0, you will have no balancing."