6

I'm not considering Etherchannel or LACP.

Given 2 Ethernet switches A and B, where both switches have port 0 configured as a tagged trunk for VLAN 100 and 101 and port 1 configured as a tagged trunk for VLAN 200 and 201. Can I connect port 0 on switch A to port 0 on switch B and port 1 on switch A to port 1 on switch B without having spanning tree shut down one of the ports or otherwise cause other problems ?

20

It depends. With IEEE 802.1Q specified STP, where you have one instance per trunk, STP will block ports.

With Cisco's PVST+ (Per-VLAN STP), every VLAN has a corresponding instance of STP. This means that if there are no loops within the VLANs, STP will not block ports.

MST, defined in IEEE 802.1s, works a bit differently. Instead of providing a one-to-one VLAN-STP mapping, VLANs can be grouped together into a few instances of STP. This way, if you do STP odd-even load balancing (odd VLANs over link A, even over link B) you only need two instances of STP, instead of one instance for every single VLAN. If you use MST correctly, your proposed configuration should work without any blocked links, just like PVST+, but with VLANs grouped into instances instead of working with each VLAN independently.

3

As others have mentioned, a per-vlan protocol will accomplish your goal. Also, you could trunk all four vlans across both links and use per-vlan costing to achieve load sharing with fault tolerance.

2

Depends on which version of spanning tree you are using, you should be able to have this work with either MST or PVST+ (Cisco proprietary).

With CST or RST, you will run into issues.

-1

Note: you'll also have to ensure any native, untagged vlan on each port is different or a loop will be created.

  • The status of a VLAN as untagged or native on a trunk has no bearing on spanning-tree operation on any platform I have worked. Could you expand on this? – YLearn May 18 '13 at 12:47
  • Depends on the flavor of STP and how the vendor coded it. The 4 vlans listed will be suitably isolated, however the unnamed native vlan (if there is one) could still cause a loop; an untagged BPDU would shutdown one of the ports. – Ricky Beam May 18 '13 at 22:49
  • Which flavor of STP are you referring to? This won't happen with STP/CST, RST, MST, or PVST+. Haven't looked at VSTP for a long time (and then only briefly), but I don't believe it will work that way either. I know I haven't come across all variants out there, so would like to know so I can avoid issues with it in the future. – YLearn May 19 '13 at 2:45
  • 1
    Think about all the vendor "extended" STP versions. My point is, ignoring the 4 tagged VLANs, the native (untagged) vlan for those ports could be a loop. This was an open question, but let's assume Cisco here... by default: native vlans are untagged, and all ports are native vlan 1. Plug port 0 to port 0 won't be a problem; then plug port 1 to port 1 and bam you. now. have. a. loop. If the native vlan is a loop, the entire port gets killed. (ask me how I know) – Ricky Beam May 21 '13 at 6:48
  • Thanks for the extra periods, they make your point so well. Referring to a loop in the network is generally considered a logical loop on the network and a bad thing. A physical loop is generally a good thing if you have ST operating correctly. With STP/CST or RST, it doesn't matter if the native VLANs are the same or not as there is only one logical topology, which will result in the scenario you provide (again whether native VLANs on the two links are the same or not). With Cisco's default PVST or a properly configured MST, only the VLAN or instance in question will be in blocking mode. – YLearn May 21 '13 at 23:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.