RSTP Topology Change Question

[Craddockc]

Community,

I am 3 weeks away from taking the CCNP SWITCH exam and had a question regarding the integrated Uplinkfast feature in RSTP. Ive learned that when an RSTP switch loses its Root Port it can immediately transition its Alternate Port into a forwarding state. In 802.1D the Alternate Port would send Dummy MCast frames to inform the upstream switches of the new path for all the MAC Addresses it had known previously on the now down port. Im reading that in RSTP, when the Alternate Port sends a TC BPDU, this serves as a notification that the MAC Addresses that were once known on the previous Root Port (that is now down) can now be reached on the new Alternate Port, so no Dummy Mcast frames are required. My question is 2 fold:

1) When the TC BPDU is sent on the Alternate Port (now the Root Port) by the downstream switch, does the upstream switch move all the MAC Addresses it once knew on the old port (that is now down) to the new port? 

2) If the answer to #1 is yes, I thought that the instant an interface goes down, the MAC addresses learned on that interface are flushed, wouldn't this create a race condition in the switch where the TC BPDU would need to be sent and received before the MACs are flushed off the port?

I actually tries to recreate this in my lab but it all happens so fast its impossible to tell. It literally takes microseconds in my lab for everything to reconverge.

Perhaps I'm not understanding this correctly, which is why I'm reaching out to the fine experts of this great Community. Thanks! 

[willwetherman · Accepted solution]

Hi,

From my understanding, when an RSTP switch root port fails and the alternative port becomes active, a TC BPDU will be sent upstream to instruct the neighbouring switch it to flush all MAC addresses in its CAM except on the port of the received TC BPDU. The TC BPDU does not notify the neighbouring switch that MAC Addresses that were once known on the now failed root port can now be reached on the new alternate port.

This is a great document that explains the STP and RSTP convergence.

http://blog.ine.com/wp-content/uploads/2010/04/understanding-stp-rstp-convergence.pdf

I hope that this helps

[Craddockc]

Thanks Will,

It would seem that the upstream switch would have to relearn all the MACs it lost on the now failed port as well as flush all the MACs from its other ports as well. In a way its confusing as to why RSTP even has a "Learning" port state at all considering how fast it brings the port into Forwarding. In 802.1D I could see the need to have a dedicated Learning state, considering how slow the networks were back then you would want to cut down on the number of unknown unicasts, but in RSTP it almost makes no sense to have a separate Learning state when Forwarding can accomplish the same thing.