We had an issue last week at a remote location. Every few minutes the IPTV system would pixilate for a few seconds - there was no significant impact to unicast traffic. PIM-SM was enabled on the L3 VLAN interface - creating the local IGMP querier - the IPTV headend was connected to a separate VLAN - with users in a "User" VLAN. We monitored the PIM state during a problem period - no change was seen - Inbound & outbound interfaces remained unchanged. It was clear this issue was at a IGMP snooping level.
Debugging for IGMP snooping showed no root cause - and IGMP state appeared to remain stable during the issue.
To cut a long story very short, the issue turned out to be related to Spanning Tree. A command making user interfaces spanning tree "edge" ports by default had somehow been removed. This resulted in all ports participating in Spanning tree, and all interfaces triggering Spanning Tree Topology Changes (TCN's) every time the interface went up and down. These TCN's resulted in Spanning Tree flushing the MAC table & flooding all traffic - with IGMP using the MAC address table for forwarding decisions, this caused IGMP to flood traffic.
So - if IGMP is configured correctly & you are getting occasional multicast floods, check whether they coincide with Spanning Tree changes!
A closer look at our Wireshark captures showed some unicast flooding at the start - but it was dwarfed by the Multicast flooding - simply due to the traffic levels from the local headend. As expected, the Spanning Tree TCN was there just before the start of the issue....... isn't hind-sight wonderful!
Just thought I'd share as looking for a multicast issue when it was an underlying Spanning Tree issue caused a lot of pain!