Configuring port-channel, disabling DTP,

[James Simpson]

Hi All

I had an issue to day where I wanted to join a 3750 x 2 stack to the network, the core is a 6500.

There was one up link configured to the core from the 3750 it had the necessary trunk configuration with BPDUFILTER enabled

at the the core end he had standard trunk configuration. The link was up and working.

Went to join a second link from the stack to the core. Decided that DTP should be disabled at the stack end SWITCHPORT NONEGOTIATE

put the normal trunk configuration on the port joined the port to the port channel. Then suddenly the logs went mad and at that point I need

Spanning-tree went mad and started blocking areas in the network to prevent loops. There is MST running in the network

I had errors similar to this  %SW_MATM-4-MACFLAP_NOTIF

I pulled the cables out of the stack and all services affected eventually came back.

I know that the BPDUFILTER enable stops the port from being involved in STP.

Please could someone confirm because of the mismatch of the config on these ports it caused this issue or is there something I missed here

Mnay thanks

[ghostinthenet · Accepted solution]

The BPDU filter is the root of the problem.

You can't have two active uplinks running without creating a loop, unless they're running in an EtherChannel/LACP/PaGP configuration. By having one or both links running with a BPDU filter, the spanning tree protocol on the 6500 didn't know that one link had to be in a blocking state and it was panicking when the MAC addresses on the 3750 stack started flapping between interfaces.

The 3750 supports cross-stack EtherChannel using LACP, so if you want the resiliency of two connections without having to factor in the spanning tree ramifications, it makes sense to bundle the two into the a channel group and have both the 6500 and 3750 treat it as a single logical connection.

If you want to keep them separated, you're going to need to remove the BPDU filters from those ports and let spanning tree figure things out properly.