10

While working on a set of switches I noticed that the spanning tree path costs aren't set up to match on opposite sides of the same physical link. For some reason that doesn't look right to me. Shouldn't the same link be set with the same path cost? Why or why not?

More specifically we have a loop configured as such:

Cisco 1/1/1 -> HP 7 A4 
Cisco 2/1/1 -> HP 8 B1 
HP 7 A2 -> HP 8 A1

HP 8 is set up with A1 path-cost 2000000 HP 7 is set up with A2 path-cost 20000

Cisco is set up with spanning-tree mst priority 0.

Tim Brigham
  • 688
  • 1
  • 4
  • 15

4 Answers4

5

According to the details you explain about your network topology, my understanding is:

  • Cisco switch is root of the spanning-tree.
  • Port A4 in switch HP 7 is root port and therefore it is in forwarding state
  • Port B1 in switch HP 8 is root port and therefore it is also in forwarding state

Now let´s focus in the other link closing the loop, the one between both HP switches. You can have only a designated port and because the cost to root advertised by both HP switches into the link is the same, switches would have to check the lowest sender bridge ID and lowest sender port ID in order to choose the designated port and the blocking port.

So, setting a different costs for ports in the link between HP switches will have no effect in choosing the designated port for this link.

Daniel Yuste Aroca
  • 2,884
  • 5
  • 24
  • 43
4

Root cost is the sum of all root port costs along the path. It is only necessary to adjust port cost on the switch where one would like to influence the root port selection. Since designated ports are not a component of the root cost, mismatched port costs on the same link makes sense.

Dennis Olvany
  • 2,435
  • 1
  • 17
  • 34
4

It's perfectly acceptable to augment the cost of a link if you want to be 100% sure of the topology that the L2 domain creates.

David Rothera
  • 2,798
  • 15
  • 20
2

The Spanning Tree Protocol specifies a switch should add the path cost of its interface receiving a BPDU to the Root Path Cost in that BPDU to learn its own Root Path Cost. It will then send out BPDUs containing this sum, ignoring the path cost of its sending interfaces.

So to answer your question, I think symmetrical costs on a link are allowed by the protocol. I can't seem to come up with a scenario in which asymmetrical costs would break Spanning Tree, or a scenario in which they would be useful for that matter.

Gerben
  • 4,690
  • 20
  • 32