Disabling irf link causes broadcasts

[pnobels]

Hi,

i just noticed following...  We have an (irf) stack of four 5940 switches.  Today i had to migrate a fibre cable of an irf link between two datacenters.  So just to be clear, we migrated from a working situation, to a working situation.  During the operation, all stack members remained up.

When unplugging one of the irf links i noticed we got a lot of broadcasts on the virtual vmware switches (14268994.276packets/s)  I can't really explain this since the four switches are still able to communicate to each other using the second irf link.   A broadcast means a device down the line has seen some reconfiguration or received a timeout.  Where the stack on layer 2 and 3 level acts as one big switch so for a device connected down the line this should not have made any difference.  The stack should take care itself of rerouting packets through another (irf) link.

Am i missing something here?  The stack also acts as a router.  Could it be the mac address of a configured ip on the stack changes?

[Vince-Whirlwind]

If you take a link down, then you lose the mac-address table associated with the interface that is no longer available. Until all those MAC addresses are re-learned on another interface, then all frames addressed to one of the MAC addresses that the switch no longer knows about will be flooded out all ports.

Are you sure they are broadcasts that you saw? A broadcast is addressed to ff.ff.ff.ff.ff.ff. Flooded traffic is addressed to a real MAC address.

[pnobels · Accepted solution]

Hi, last friday i performed the same operation on the second irf link.  I received no such alarms.  After querying around i found out that during the first irf migration, a collegue was at the same time performing a bunch of vmware vmotions.  This caused the broadcasts.  Case closed.