[c-nsp] IGMP differences between c3560 & c4948

[Christopher.Marget at usc-bt.com]

Said Vince:
> On the C3560, with default config, i believe igmp snooping is enabled
> by default.  A multicast source floods all traffic to all ports in the
> vlan, until I configure 'ip igmp snooping querier' at which point the
> flooding stops.  I would have expected igmp snooping to figure out that no
> receiver requested traffic, so no flooding should never occur at all ?

The behavior you describe sounds correct.  IGMP snooping depends on IGMP traffic in order to have something to snoop.  Without an IGMP querier, you can only count on the initial (unsolicited) IGMP host report.  There might be no other traffic, rendering IGMP snooping is totally helpless.  It makes sense that traffic would flood to all ports until you enable a querier.  This sounds like a feature:  The switch is smart enough to disable multicast filtering when the information required for good decision-making isn't available.

> If I remove 'ip igmp snooping querier', and configure 'ip pim sparse-
> mode' on an SVI in that vlan, again I would expect the 3560 to stop the
> flooding, but the flooding goes on an on, which i dont expect.

Is 'ip multicast routing' enabled?  What does 'sho ip igmp snooping querier' have to say about it?

> On the 4948, flooding again does occurs with default config with igmp
> snooping enabled by default, but different to the c3560 both 'ip igmp
> snooping querier' and/or 'ip pim sparse-mode' take care of controlling
> that.

Last time I needed 'ip igmp snooping querier', it wasn't available on the 4948.  What version are you running?

> am i getting it wrong, or is it indeed by design that the boxes behave
> differently ?

It could be different defaults (or manual configuration) around enabling L3 multicast.  Without 'ip multicast routing', 'ip pim sparse=mode' is a no-op.

Good luck.

/chris