A server setup relying on multicast on 18.104.22.168/ udp 45591 (internally on the vlan - non-routed) was moved from one vlan on a Catalyst 2960 switch (working OK here) to a datacenter vlan.
First it was established on 4 gigabit ports (fex) on a Nexus 5010 setup: apparantly 22.214.171.124 was not forwarded by the switch (wireshark check on both source and destinations), but f.ex. 126.96.36.199 was forwarded for a period of time - then it stopped, apparantly blocked by the switch ! We tried to move the servers to 4 gigabit ports on a Cat65 module with same result. Eventually the application guys changed the multicast IP to 188.8.131.52/ udp 45591 and everything works fine !!!
Is there something with switches and multicast 184.108.40.206/24 ?
23 bits to the right should have been yellow in all 4 lines!
L3 to L2 mapping: Multiclass Class-D always starts with 1110. Hereafter 32-4= 28 bits left. You only have 23 bits for multicast L2 addresses, that is 28-23 = 5 bits ovelap - 23 bits marked with yellow. 224.0.0.x is Link Local Multicast and will always be flooded in switches. 220.127.116.11 maps to the same L2 address range (0x0100.5e00.00xx) as the Link Local Multicast, so eventually it will be flooded as well with as without igmp snooping enabled. 228.10.10.x will not be L2 flooded with igmp snooping enabled (=default).