Multicast Flood - Explanation

[simoesmarco8626982]

Hi all,

has a weird situation in one of my installations configured for multicast where when a computer would shutdown or reboot the switches would flood all the ports with multicast traffic and then stop after a while. I investigated this and discovered that by putting the command "no ip igmp snooping tcn flood" the flood would stop. 

For what I saw this command prevents that if there's a topology change the switches won't flood the ports with multicast data and the groups that are already active stay active.

But my question is, I have several sites configured with multicast (PIM Sparse Mode with RP or Auto-RP configured) in the same fashion as this one and none of the other sites ever had this issue and by consequence never had to configure this command. Is it possible to explain please why this is the case? I would like to learn the reason

Thank you

[Elliot Dierksen · Accepted solution]

The problem isn't the multicast itself. the problem is the "TCN". That means a spanning tree topology change notification. If that is an edge port (as in one facing a single host), it should have "spanning-tree portfast" on that switch port. The most widely recognized reason for this is that without portfast, DHCP usually fails. The other thing that portfast does is to tell the switch to NOT send a topology change notification when the port goes up or down. I would strongly suggest that that you also enable "spanning-tree portfast bpduguard default" globally. That will cause edge port with portfast enabled to go into an err-disable state if they receive a BPDU. You only want portfast on edge ports, NOT links between switches. If a link between switches goes down, that should generate a TCN. BPDU guard will disable a single port to save the rest of your network from a loop, and that is a good thing.

[simoesmarco8626982]

Thank you very much for you explanation @Elliot Dierksen , I will do this changes in the switches