rapid spanning-tree path cost method

m.goretti · ‎02-20-2008

I've a problem in my network infrastructure. In the backbone area there are a couple of Cisco Catalyst 4510R, the primary and secondary root bridge of the network. The spanning-tree protocol used is Rapid Spanning-tree with a pathcost method short (default) . The distribution switches are stack of 4/6 Linksys SWR224P connected to 4510Rs in a ring topology. These switches used a pathcost method long.

Sometimes the Catalysts freeze and I see

C4K_EBM%-4-HOSTFLAPPING: on the log.

Surely it is a problem of the loop, but I do not know what could depend.

The network works, then for some cause yet to be discovered, the Catalyst

seems to stop.

It can be a mismatch of pathcost that after an event not able to calculate the correct path of spanning-tree?

Francois Tallet · ‎02-20-2008

Hi,

The mismatch in the stp cost method will just make the final topology more complex for us human to understand, but it will still be deterministic. In fact, the cost has always been encoded on 32 bits in the BPDUs. It's just that before the long path cost method existed, only 16 bits were used, now the full 32 bits are configurable. STP does not care about that and always computed its topology based on 32 bits.

I would not think your problems are linked to this inconsistency in the path cost method, unless your network has a huge diameter and the location of the root bridge is a the edge (BPDU expiring because of a too large diameter). That's relatively unlikely.

I used to work in the TAC and the first question I would have asked is: When did the problem occur? Is it a new deployment or a problem coming from a particular change in the network? ;-)

Regards,

Francois