Description of problem:
When igmp snooping is enabled and VMs are running on 2 different compute nodes multicast traffic sent by one VM is not reaching second VM (subscribed to the relevant multicast group).     

Version-Release number of selected component (if applicable):
RHOS-16.1-RHEL-8-20201214.n.3

How reproducible:
100%

Steps to Reproduce:
Note, use environment with ML2/OVS mechanism driver.
1. Make sure igmp snooping is enabled in neutron and br-int on compute nodes.
In order to check you can run "sudo podman exec -it neutron_api crudini --get /etc/neutron/neutron.conf   ovs igmp_snooping_enable" on controller nodes. It should return True. And also "sudo ovs-vsctl get  Bridge br-int mcast_snooping_enable" and "sudo ovs-vsctl get  Bridge br-int other_config:mcast-snooping-disable-flood-unregistered" return 'true'

2. Create router, tenant network/subnet, security group allowing icmp, ssh, igmp and udp port 5001 (we will use it). Launch 2 VMs on the network. Make sure both VMs are running on different compute nodes. 
3. Subscribe one VM (receiver) to a multicast group (note: the group should not be from range 224.0.0.X, e.g. 225.0.0.100 is OK)
4. From the second VM (sender) send udp packets to the multicast group (use same port as security group allows, e.g. 5001).

Actual results:
Packets do not reach the receiver VM.

Expected results:
Packets reach the receiver VM

Additional info:
Note, in case both VMs are running on the same compute node multicast traffic from the sender reaches the receiver VM successfully.