RSTP - convergence

[bapatsubodh]

Hi,

In the rstp convergance process, if any switch do not get BPDU from it's neighbor on the root port then it consideres that neighbor as dead. Then sends it's own BPDU as the root BPDU on the designated port negotiates with the neighbouring switch and will decide whether the existing root is still working or not and then will put the port in to root port. ( considering the existing root switch is working fine)

Please consider the diagram ( sorry it's not perfect image) if the link between switch 1 and 3 is broken. Then as per the procedure, switch 3 need to wait for three BPDU period that is ( 6 second) then negotiate with switch two. Switch two ALT port and switch three DP will negotiate that the existing root that is switch 1 is still working fine and ALT port will move into DP for that segment next existing DP port of switch 3 will move in to RP. After that switch 3 will initiate TC BPDU those will be flooded all across the network this flooding will take place as long as switch 3 is keeping TC flag set. All switches will flush the MAC  tables and rebuild them.

If this is accurate then how does the RSTP converges so fast. It should wait at least for three BPDU time before it starts negotiating with switch 2 (in our case). There is absolutely no ping packet loss when we disconnected the cable between sw1 and sw3.

Please share the experience.

Thanks

Subodh

[Mohamed Sobair · Accepted solution]

Hi Subodh,

In breif, this is the advantage of running RSTP,

RSTP already should have previously calculated the following:

1- A quick converged (backup port) for the root port.

2- A quick converged (Alt port) for the designated port.

So, as soon as the link between SW1 and SW3 goes down and after SW3 waits for the 6 Sec dead interval, it sends inferior BPDU out its DP port to SW2, SW2 drop this BPDU as it has superior BPDU already learned from SW1 and forward it superior BPDU out its designated port (which was previously Alt Port)  to SW3 Root port.

However, there should be about sub seconds for this convergence to happen,

HTH

Mohamed

[Giuseppe Larosa · Accepted solution]

Hello Subodh,

disconnecting the cable causes physical layer to goes down

if you want to see the 3 BPDU hello interval in action you should apply spanning-tree bpdu-filter enable on sw1's port to SW3 in this case the link would stay up but the root port would not receive BPDUs from root bridge.

Hope to help

Giuseppe

[bapatsubodh]

Hi Giulsar

Thanks for your exact pin point answer. I didnt find this in any of the documentation.

subodh