What would cause this - odd STP issue.

[mhoppes]

Another bizarre netonix issue.

RSTP enabled on the ports in question.

I don't even have an explanation for how this could be configured to work this way.

Switch A - port 1 AF24. Port 2 AF11
Switch B - port 1 AF24
Switch C - port 1 AF11 Port 2 connects B and C together.

The AF11 carries the traffic normally. However, when the AF24 link drops the AF11 stops passing traffic until the 24 comes back.

[Stephen]

Do you have the logs of the event?

[mhoppes]

I did, but cleared them as I didn't see anything that showed anything meaningful. We have a maintenance window scheduled for Thursday morning when I'm hoping to try some things. I'm just trying to figure out under what conditions this type of things could even possibly happen where the second side goes down and takes down the first side when it does, which had previously be passing traffic properly.

[Stephen]

Hmmm, well if RSTP was responsible, maybe the root switch had to be re-established. If that happened than it could temporarily bring down the other link while it re-converged and re-calculated the available root on the network.

Just a guess though.

[mhoppes]

I've seen it go down for 15 minutes at a time though. I'll let you know what I find out.

[mhoppes]

So it seems this was caused by Switch D, which hung off of Switch C. Somewhere in history Switch D became the root switch. So when the link would go down, no body could figure out where the root switch was.

But this still doesn't explain, to me, why it would cause the secondary link to go down when the primary went down. Nothing shows exciting in logs. But after moving the Root Switch to A the issue has now cleared.

[Stephen]

I figured it was some variant of that.

This link might prove helpful for others in the future:
https://www.auvik.com/franklyit/blog/sp ... -mistakes/

In section 2 "Letting the Network pick your root bridge" they describe a scenario of how this can occur.
Basically it's an unfortunate limitation in the Spanning Tree Algorithm that it cannot converge correctly when finding the root bridge in every topology.

Your solution for manually adjusting which switch became the root in the network to my knowledge is the best answer.

[mhoppes]

I get that... but it almost seems like the Netonix switches don't reselect a new root bridge when the root bridge vanished from the network?