NOTE: I came across this issue at my work, and have since built a small lab to isolate the problem.
Hi everyone!
I have three EX4200s (We'll call them A, B, C) running 12.3R12.4 with basically a stock configuration. They're connected in a loop, all vlans trunked, with MSTP running (all vlans in the CIST). I configured the port cost higher on switch C, so that the port connecting over to switch B would block. Everything works fine just as you'd expect.
Now, I connect a Cisco Catalyst 3650 to switch A and trunk all the vlans over. The Catalyst runs rapid pvst+ on all VLANs. For years this ran just fine, the two different STPs not interfering with the other. This works perfectly fine in my test environment.
Now, part of some routine maintenance, I swapped out switch C for a 4300, configured the MSTP port cost over to switch B the same way as the old one. The port blocks just fine, and is labeled as an alternate path just like before. All is well!
Here's the problem: on the 4300, even though the port is in MSTP block mode, the rapid pvst+ traffic from the Cisco is still passing though the blocked interface. This means the rapid pvst+ traffic is looping continuously, and going back to the Cisco. The Cisco immediately blocks all vlans on the link over to switch A.
The only fix I can find for this is to instead set the port cost on one of the other 4200s, so that the trunk interface is blocked on the 4200 side and not the 4300 side. The 4200 does not let the rapid pvst+ packets to traverse the blocked interface, but the 4300 does.
Any help would be much appreciated.