PIM Multicast - 2 sources, 2 paths, 1 or 2 dest. group = RP Problem!

response3 · ‎10-12-2007

I have a somewhat unique application that uses Multicast for dozens of client devices to send data to a small group of servers.

The client devices use 239.192.1.1 as the destination IP, and the servers are members of this group. The servers are connected to a layer-2 switch. This switch connects to 2 switches (Cisco Catalyst 6500 and 3560) running IP routing and PIM sparse-mode Each router has an RP set. These switches each have a unique subnet behind them, with these client devices connected.

The problem I need help with is this:

When I only have the clients on switch A powered on, multicast messages flow downstream towards the servers, through the layer 3 switch. When I power on the second group of devices on switch B, there seems to be an RP designated router 'election' to the switch with the higher IP on the server vlan (in this case, the 3560E). Once this happens, the group of devices on switch A can no longer get their multicast messages forwarded thru switch A, only devices on switch B are able to do this. If I shut down switch B, the change is not reversed until I shutdown the both interfaces (vlan101, 111) on the 3560 and reboot the devices on switch A. This puts it back to the original state with clients on switch A communicating, and switch B shutdown. Bringing the interfaces on the 3560 up repeats the problem.

We also tried getting the clients on switch B to send to a different group (239.192.1.3), that didn't work either, so there must be some problem relating to the RP, I think.

What I'm not sure of is if there is something I can do to keep both groups of client devices 'online' and sending messages to the same Multicast group via two separate paths, or if that is even my problem. Any ideas? Thank you very much.

Kevin Dorrell · ‎10-12-2007

I have a couple of questions:

Look at the IGMP snooping on the switch. Is it seeing both router ports as mrouter ports?

"Each router has an RP set." Is the RP one of the routers? Which one? Could you explain or post your RP configuration? Normally you would have only one RP IP address, otherwise you have multiple shared tress. This would reside one one router only, unless you want RP redundancy. If you have RP redundancy, then you need to run MSDP between them.

Kevin Dorrell

Luxembourg

konigl · ‎10-19-2007

I think it's an RP issue. PIM-SM wants to see one RP per multicast group. You appear to have two. How are you specifying the RPs on your routers now?

You CAN have two (or more) in the same PIM-SM domain, if you implement "Anycast RP". This uses MSDP between the RPs, what Kevin mentioned.

In the information you posted, it looks like 192.168.1.1 and 192.168.2.1 could be the unique loopback IP addresses to use in the MSDP peering part of an Anycast RP config. Then, create another loopback interface on each router that uses the same IP address on each router: for example, 192.168.0.1. And specify THAT shared address statically as the ip pim rp-address on each router. If you're only doing this on two routers, you're done. (If three or more are going to be RPs in an Anycast RP scenario, then you also need to configure MSDP mesh-groups.)

Otherwise, I would recommend picking one router to be your RP (either your 6506E, or your 3560E), and statically configuring its IP address as the ip pim rp-address on both your routers.

Here's a couple of links that explain Anycast RP in more detail:

Configuring a Rendezvous Point

http://www.cisco.com/en/US/tech/tk828/technologies_white_paper09186a00800d6b63.shtml

Anycast RP

http://www.cisco.com/en/US/tech/tk828/technologies_white_paper09186a00800d6b60.shtml

Hope this helps.