spanning-tree and unmanaged switches loop cable problem

[willyard]

I have a question regarding spanning-tree and unmanaged switches. We use a core switch that supports spanning-tree and is managed. From this switch we have fiber optic connections running to each classroom (star topology). At the classroom end we use an unmanaged switch (no stp) which connects to all of the wall plates in the classroom. We have a problem of students and teachers disconnecting the network cables and moving computers. When they do this they will sometimes loop a cable from one wall plate to another causing a broadcast storm that takes down the entire school.

If I turn Spanning-tree on on my core switch will it take care of this problem even though the unmanaged switches do not support spanning-tree. I don't mind if the offending classroom gets disconnected, but I would like school to stay up.

[Kevin Dorrell]

I would enable Spanning-tree (It is enabled by default), and put portfast on all the ports of the core switch that connect to the unmanaged switches. Then configure bpdu-guard on all the ports. The ports will be generating bridge BPDUs, even if they are in portfast. But if the port sees a BPDU looped back, it will disable the port and save the rest of your network. But do NOT configure bpdu-filter as that will stop the port generating the BPDUs in the first place, which will disable the safety mechanism.

Here it is for CatOS:

http://www.cisco.com/en/US/products/hw/switches/ps4324/products_command_reference_chapter09186a00801f61ed.html#wp1072868

http://www.cisco.com/en/US/products/hw/switches/ps4324/products_command_reference_chapter09186a00801f61ed.html#wp1049757

Here it is for IOS:

http://www.cisco.com/en/US/products/hw/switches/ps4324/products_command_reference_chapter09186a008037a086.html#wp1065041

http://www.cisco.com/en/US/products/hw/switches/ps4324/products_command_reference_chapter09186a008037a086.html#wp1076298

I just tried all this in the lab, and it does do what you are proposing. As soon as I put the loop on the non-SPT switch, the "core" switch err-disabled the uplink port.

However, I found out something else that I am still investigating: that without the bpdu-guard, my "core" switch put the port into a spanning-tree state "LBK" which I wasn't previously aware of. I have to investigate what this state does, when it was introduced, and whether it would automatically protect the rest of the network. When I removed the loop from the unmanaged switch, it took 20 seconds to come out of the LBK state, and wnt through LIS and LRN, even though I had portfast enabled. But it does mean that the network recovers automatically.

To generate this loop, I had to use a crossover cable on the non-SPT switch. This means, I guess, that either you have crossover cables available in the classroom, or your dumb switches are auto-polarity.

Hope this helps.

Kevin Dorrell

Luxembourg