I was brought in to troubleshoot an interesting issue with some Cisco switches yesterday. I had a stack of 3750′s(Core) and a stack of 2960S’s. Etherchannel was set up on both sides to connect the two stack’s together. Each time they were connected they would work for about 20-30 minutes and then bring the network down. i immediately suspected Spanning Tree to be at fault but after some investigation with a colleague we found that Spanning tree wasn’t the issue.
In troubleshooting we discovered that when we plugged in the port channel there was immediately a massive amount of data transversing it, far too much to be normal. Doing some digging we discovered there was some mirror port configuration on the switch which was mirroring one port of my Etherchannel to another port on the Etherchannel thus causing a type of traffic loop which eventually completely floods the switch.
Not likely an issue I will see again but a valuable lesson on pruning old configurations after they no longer apply.