Multicast RP not in sync

[dominic.caron]

Hi

My security camera are working on top of my IP network using multicast for live feed. My problem is that some camera, after a while, say 1 or 2 weeks, need to be restarted.

I ve traced back the problem to an out of sync multicast entry. The pim DR router on the subnet is receiving the flow and as a entry in is mroute. The RP, on of my core router doesn't. To fix this, I have to clear all mroute in my distribution so it re-create them on the Core RP.

Any idea?

[mikedavi1]

Hi Dominic,

A few questions for clarification that could help the forum with troubleshooting:

Are you saying that the RP completely looses the (*,G) mroute

Or both the (*,G) and the (S,G)?

Do you see the S,G anywhere besides the receiver's DR during a failure?

Are you using PIM v1 or V2?

Are you using IGMP v1 or V2?

Are you saying that simply clearing the mroutes restart the camera function? Or do you need to restart the camera as well?

Are there any interfaces between the camera and RP that are not pim enabled?

Not sure what you mean when you say the receiver router is receiving the flow for the "broken" camera. Do you mean there is a partial mroute (missing incoming interfaces)?

Or is that mroute complete with both incoming and outgoing interfaces & traffic counts incrementing (show ip mroute count)?

Cause if the camera needs to be restarted, what is that traffic flowing across the router?

Some thoughts:

It would be normal for the RP to lose the S,G as it is typically pruned upon the shortest-path-tree (SPT) switchover. At that point, the RP has no need for the specific S,G since the reciever's router switches from the shared (RP) tree to the SPT.

However, periodic registration messages from the camera's router should keep the *,G entry intact. Refreshed S,G entries only should occur upon new joins from receivers.

Also, default timeoutes for mroutes are on the order of 3 minutes, not 1 week. Of course these can be tuned. You can verify in the show ip mroute command.

It possible that dynamic routing changes caused by some event are moving traffic away from the pim interfaces causing RPF failures?

If you are losing the *,G on the RP, I'm imagining a few possibilities:

1) it sounds like PIM register messages are not making it from the camera

2) register messages are making it, but RPF performed by the RP is failing intermittently

3) register messages are not being sent by the camera's router/DR

4) the camera stops sending multicast traffic

5) igmp is failing between the camera and the camera's DR.

6) igmp snooping is broken between the camera and the camera's DR

Mike