We are operating several Cisco 2960S stacks here and, for some time, are experiencing regular Spanning Tree TCNs (Topology Change Notifications).
When using show spanning-tree detail to get to the source of the TCN, they originate from different 2960S stack members.
As the config has been transferred from older switches, the ports are configured as trunks, with the data VLAN being the native VLAN and the VoIP VLANs being included as tagged in the trunk. I know, today the correct configuration would be switchport mode access combined with switchport voice vlan - but at the moment it is at it is.
Part of the problem was based on the fact that some ports were only configured as portfast, not portfast trunk, so this has to be fixed, as the portfast only command is not applied to ports in trunk mode. However, some ports are configured as portfast trunk, which as of my understanding of Cisco documentation, should avoid generating TCNs on link changes.
Sadly, they do.
Executing show spanning-tree interface Gi1/0/x portfast shows portfast enabled for all VLANs on that interface, but show spanning-tree detail together with show logging shows that these interfaces are indeed the source for the Spanning Tree changes.
It is made sure that there are no "special" devices connected to those ports, just VoIP phones and desktop workstations.
The version run is 12.2(55)SE3, the devices are 2960S-48LPS-L and 2960S-48FPS-L.
Looking for known bugs, there was a bug in a 12.1 release for 3550 switches that caused such behaviour, but this is the wrong device and a much older version, and I guess somebody should have experienced that bug before.
Any idea where to look, what to try?