Spanning Tree Topology Change Notification (TCN)


(Rene Molenaar) #1

This topic is to discuss the following lesson:


(Yevgeniy O) #2

you’re strong man :wink:


(Alberto D) #3

There is something that i not understand . I worked with network with many hosts , but I didn’t set these in portfast . However there wasn’t problem like you mentioned (burned traffic) , why ?


(Rene Molenaar) #4

Hi Alberto,

It’s a good practice to disable TCN generation for interfaces that are connected to hosts. An interface that connects to a host/server doesn’t change the STP topology so there’s no reason to generate TCNs when these interfaces go down. It’s possible that you never encountered the downside of not enabling portfast but an example like the problem with the backup server is possible.

Rene


(Atif S) #5

Little confused towards the end, if portfast is enabled on host end and when host interface goes down how does the mac address get cleared on the switches since the tcn is not sent on portfast ports? after the default 5min age time?


(Rene Molenaar) #6

Hi Atif,

TCNs are only for spanning-tree. When an interface goes down on the switch, it will clear the MAC addresses associated with it.

Rene


(Mauro P) #7

Hello Rene,

I think now is possible to configure (spanning-tree portfast trunk) for ports in trunk mode. I found that amazing. :slight_smile:

Regards Mauro.


(Rene Molenaar) #8

Hi Mauro,

Yes that’s right, it is possible. This can be useful in certain scenarios…for example standalone access points that require a trunk because they use a different VLAN for each SSID.

Rene


(Gbenga R) #9

thank you for the lesson. it definitely cleared a lot of doubts.


(Chandima E) #10

Hi Rene

It is a great lesson to understand the Topology change on the STP. I have a question which is related to the switches which receives only BPDU with TCN flag set from the Root bridge

Assume that Switch network has few more switches and TCN notification generated by switch ( switch A on the diagram) has not been pass through on these switches . I need to know what would be the mac address aging time for the other switches which receives BPDU’s with TCN flag set. My understanding is if a switch in the switch network received the TCN from the Root bridge the switch set the mac address aging time to max-age in the STP.

My other question is, due to some reason when switch does not receive the BPDU, will this switch wait for Max age time (20s) by default to flush the Mac-address table? Then how the STP reconvergence will affect the switch network

Regards

Chandi


(Davis W) #11

Hi Rene,

By default the Aging time 5 minutes will clear all the Mac address in the table and relearn it or only clear those host MAC address that have no activity for 5 minutes?

Davis


(Rene Molenaar) #12

Hi Davis,

The aging time applies only to specific MAC addresses. Wiping the entire MAC address table every 5 minutes would be a bad idea :slight_smile:

Rene


(Davis W) #13

Ok. Thanks Rene.

Davis


(shayan a) #14

Hello Rene,

I am getting confused with this concepts in STP, The more i read the more i am getting confused
Please help me upnderstand this concepts, I 've wriiten my understanding and questions.

In a Steady state Network assuming the packet between computer A and Computer B takes (Switch B --> Switch A --> Switch C --> Switch D)

**Question related to interface states and convergence time**

Switch C (Root Bridge) generates configuration BPDU’s and sends it to Switch A and Switch D and in turn they relay it to switch B
When we have an indirect link failure (shutting down the interface between Switch A and Switch C i.e fa0/16)
Switch A will start advertising itself as Root switch to switch B (downstream)
Since the BPDU received at switch B is inferior to the BPDU which it had received earlier, it will start the max-age timer ignoring BPDU’s till 20 secs
Once the max-age is expired the interface fa0/19 of switch B will move into Listening state, eventually move out to learning and forwarding
Totally it takes 50 sec for the interface (fa0/19) to move from blocking to Forwarding state
what exactly happens in the listening state ? it will start electing the root switch, root port, dp etc ? is my other understanding are correct ?

**Question related to Topology Change**

When ever we have a topology change the switch with interface going up/down, will generate a BPDU with TCN flag set and sends it towards the root switch
This is the only time where a non-root switch will generate a BPDU towards the root switch
The switches which receive the BPDU will acknowledge back to the sender switch which generated the TCN and also relay the BPDU towards the root switch
once it reaches the root switch, it will propagate the changes to the entire switch network in that segment
upon receiving on this BPDU every switch in that segment will reduce mag-aging timer to 15 secs
so my question is TCN is more of to do with aging out the mac-table and it has no significance in the interface state changes ?

I’ve written my understaning and questions. Pleae help me with this questions.

Regards!
Shayan


(shayan a) #15

Hello Rene,

can you please shed some light on this ? The more I explore the more i am getting confused.

Regards!
Shayan


(Rene Molenaar) #16

Hi Shayan,

If you want to see exactly what is going on in a topology like this I highly recommend to configure 4 switches and to enable “debug spanning-tree events” on all switches. It will show you exactly why and how the interfaces are moving through the different states. It’s much easier to see it then just to read and visualize how it works.

In the listening state we only process BPDUs. This time is used to learn what the STP topology looks like, it’s when we elect ports etc yes.

The main reason for the TCN is to reduce the MAC address table aging timer from 300 to 15 seconds. If we don’t do this then it’s possible that traffic gets blackholed for up to 300 seconds.

Rene


(shayan a) #17

Thanks Rene,

I did try with the topology and got my doubts cleared. but just wondering in your reply you wanted to say in listening it will send BPDU’s ?
I need to try the same for RSTP now since i am getting confused :slight_smile:

“In the listening state we only process BPDUs. This time is used to learn what the STP topology looks like, it’s when we elect ports etc yes.”

Regards!
Shayan


(Andrew P) #18

Shayan,

You are correct. In the STP Listening state, the router will both send and receive BPDUs, but it will not forward any other network traffic.

--Andrew


(francesco r) #19

Hello Rene,
i would like to see if I understand theory Of tcn:
When a non-root switch detect a topology change ,maybe an interface goes down, it delete mac learned From that interface , generate and forward a tcn towards root bridge on its root port and reduce aging time to forward delay.The upstream switch detect on its designated port this tcn, ack this tcn, reduce its aging time, generate a tcn itself and forward to root bridge on its root port, right? These 2 switch has not to wait tc of rootbridge, to reduce aging time, maybe other switch in other side Of the network, reduce aging time as they receive tc generate by rootbridge… Is correct my discuss?
best regard


(Rafa) #20

Hello Rene,

In addition to the mac address table being cleared faster upon receiving TCN’s, does it also clear the ARP entries (faster) for it as well?
Or is the faster clearing, strictly mac address table?