I've been chasing this issue for a couple of weeks now, but I'm running out of ideas and/or not understanding how to have this multicast network setup, so needing some help.
We have a vCenter environment with 4 ESXi hosts. These hosts have a DSwitch setup for Multicast, and this goes to a single VLAN (200). One VM (10.10.12.4) in this environment produces the video feeds over the Multicast VLAN. We have run into an issue recently where all feeds are only making it to VMs within the same ESXi host. If there is a VM trying to access those feeds outside of the same ESXi host, then the results are sporadic. Some get one or two feeds (from different IP ranges), some don't get any feeds at all. If we move the feeds VM to another ESXi host, the same behavior happens. The multicast link from this vCenter goes to a Nexus C9348GC-FXP breakout port 4x10G to distribute to other devices within our network on the same VLAN. The vCenter Multicast DSwitch connects to the breakout port over a trunk for VLAN 200. In vCenter, the IGMP Snooping is enabled, and the Port Groups are set for VLAN 200. This was working up until a certain point. We have storm control level 95.00 on the Nexus ports for Multicast, but metrics don't show this as a problem (unless I'm looking at the wrong metrics).
We have another vCenter environment that is connected to the same Nexus switch as the feeds source, but most VM's in the second vCenter environment are not getting traffic from the feeds source. The same applies to servers connecting from other Cisco switches (a Nexus and an SG500X with servers connected) trying to pull from the same source over VLAN 200.
We have:
- vCenter 1 ESXi 1 containing Feeds VM -> Multicast dsSwitch on VLAN 200 -> N9K-C9348GC-FXP (Core Switch) VLAN 200
- vCenter 2 with 1 ESXi host with a regular Multicast vSwitch on VLAN 200 connected to same Nexus C9348GC-FXP on VLAN 200 getting 0 feeds.
- A test environment with a few baremetal servers connected to a N9K-C93108TC-EX that has a trunk to the core N9K-C9348GC-FXP. The servers have their own access ports on the N9K-C93108TC-EX to receive multicast traffic from VLAN200. At one point, these servers were getting the Multicast traffic just fine, but it now gets some of the feeds, not all. Similar to the behavior of the VM's of vCenter 1 not on the same host as the Feed VM
I'm seeing mixed results in my searches that it can be IGMP/PIM settings or vCenter issues, but would like to have someone look at the configs as a sanity check, or give me some debugging steps to try and narrow this issue down.
For the N9K-C9348GC-FXP (Core hosting the 2 vCenters):
IGMP:
vlan configuration 200
ip igmp snooping fast-leave
interface Vlan200
ip igmp version 3
PIM:
feature pim
ip pim rp-address 10.10.12.1 group-list 224.4.0.0/16
ip pim rp-address 10.10.12.1 group-list 226.6.0.0/16
ip pim rp-address 10.10.12.1 group-list 227.7.0.0/16
ip pim rp-address 10.10.12.1 group-list 228.8.0.0/16
ip pim auto-rp mapping-agent Vlan200
ip pim ssm range 232.0.0.0/8
ip pim auto-rp forward listen
VLAN200:
interface Vlan200
no shutdown
ip address 10.10.12.1/24
ip pim sparse-mode
ip igmp version 3
vlan configuration 200
ip igmp snooping fast-leave
For the N9K-C93108TC-EX (hosting the baremetals):
IGMP:
vlan configuration 200
ip igmp snooping fast-leave
interface Vlan200
ip igmp version 3
PIM:
feature pim
ip pim rp-address 10.10.12.1 group-list 224.4.0.0/16
ip pim rp-address 10.10.12.1 group-list 226.6.0.0/16
ip pim rp-address 10.10.12.1 group-list 227.7.0.0/16
ip pim rp-address 10.10.12.1 group-list 228.8.0.0/16
ip pim ssm range 232.0.0.0/8
ip pim auto-rp listen
interface Vlan200
ip pim sparse-mode
VLAN 200:
vlan configuration 200
ip igmp snooping fast-leave
interface Vlan200
no shutdown
ip address 10.10.12.2/24
ip pim sparse-mode
ip igmp version 3
I think if I can get the kinks worked out with vCenter 1 and vCenter 2 on the Core Nexus switch, I should be able to figure it out on the neighboring switches, but just needing help on what to look for or what I may need to change in the configuration to test.
Thanks
