spanning-tree interopability problems with cisco

[Carl Boyd]

Here's my scenario:

Switch A is a cisco 3550, running an 802.1q trunk to switch B, which is a Procurve 2524. The Procurve (switch B) has a port set to vlan 95 (untagged/native). Switch B connects back to another cisco switch (switch C) that I have no control over.

The problem is this: switch C is seeing 802.1Q BPDUs coming thru the Procurve switch (which I actually think are originating from switch A, not the Procurve), and switch C disables the port because the BPDU types do not match. I can't run a 802.1Q trunk between switch B and C (like I said, switch C is out of my control). The only solution to this that I've found thus far is to disable spanning-tree altogether for vlan 95 on switch A. I can leave spanning-tree up for all the other VLANs and everything is fine.

I know what the problem really traces back to, that would be cisco's insistance of using PVST, without the option to revert to the actual 802.1D standard. My question is this: is there any way I can set the Procurve to not pass along the BPDUs coming out of switch A, or else a way to "force" the port to a non-trunking mode (ala the cisco command "switchport mode access")? I know I could just leave spanning tree disabled for VLAN 95 (there is literally only the one port in that VLAN, it's only used for an uplink outside the network), but since I don't have this problem in an all-cisco environment, I'd like to figure out some other solution.

I was thinking of maybe playing around with using MST instead of 802.1D, but the 2500-series Procurves don't appear to support it (only the 2600s and higher), plus then there's the issue of older cisco's (2900xl's for instance) not supporting it either.

Of course, what I'd REALLY like is for $(*@#@@!!'ing cisco to just adhere to the standard, but we know that won't be happening.

Does anybody have any ideas on this?

[Bruce Campbell_3]

I have seen two spanning tree ideosynchrasies
when mixing Cisco and another vendor (it was
Extreme when I ran into this).

Cisco switch sends the BPDU's on trunks
SNAP encapsulated. The Extreme switch
that received them did not un-SNAP
them, and forwared them as usual traffic.
Any Cisco switches on untagged ports then
exitted the network with a message about
"802.1q BPDU received on non trunk".
Extreme eventually dealt with this.

The other problem is when connecting
untagged ports between Cisco (and other)
switches running CDP (Cisco Discovery
Protocol). If the native vlan at both
ends is not the same, CDP disables the
port, and logs a message. Solution is to ensure connections between untagged ports
always use same native vlan, or always
use 802.1q between switches, or turn cdp
off.

I think if you always use 802.1q between
switches, and never use native vlans, you will be OK.