[c-nsp] spanning-tree problem, two roots?

[matthew zeier]

For reference, http://www.velvet.org/~mrz/stp-two-root.png


The 6509s are all connected to core1s and core2s on dot1q trunks.  The two 
core switches are also connected to each other.

The new installation I'm working on involves the 2970s and 2950s.  The 2970s 
are single homed off a 6905 running pvst+.  The 2950s are multihomed on each 
of the 2970s.  Due the limit of STP instances a 2970 can run in pvst, I 
changed the mode to mst (because I wanted to continue running vtp and we 
have more than 240 some vlans).

When a 2950 is connected to both 2970s, both trunk ports show FWD and some 
servers in the server cloud are unreachable.  If I shutdown one trunk port, 
those servers become reachable again and stay reachable while the trunk is 
in BLK and LRN.  Often the 2950 appears to get into a confused state and all 
devices are unreachable until I reload the switch (this is often after a lot 
of shut/no shuts on the trunks).

How do I resolve this?  Do I need to goto pvst?  I suspect the 2950s think 
they are roots but that's not the desired setup (and it's not working that 
way either).

Any help is most appreciated.