How to force receiver router to use desired RP if anycast RP configured?

[jjia]

Hi,

As per document http://www.cisco.com/en/US/tech/tk828/technologies_white_paper09186a00800d6b60.shtml,

if we configured anycast RP in multicast environment, Sources and receivers will use closest RP.

Is it possible for me to force the receivers to use one specific RP as primary RP even this RP may not be the closet RP?

and how?

Can we use some thing like access-list on the receiver router to force use one RP instead of the other RP? Is there any example for using access-list to achieve this?

Other option is try to tune the metric so that the closet RP will be the desired one?

Thanks,

JJ

[Harold Ritter]

If you want to be able to determine which RP a given multicast router will pick then Anycast RP is not for you.

Can I ask you why you care about which RP will be selected for a given multicast router.

Hope this helps,

[jjia]

Hi,

This is the secenario:

Customer has site A(HQ),B(remote Backup),C,D(both remote sites). They use MPLS for wan connection. ON Site A,and B has same multicast application which are on the same multicast group (the application supports sparse mode only but will not support application redundancy) sources for A, B, C, D site users. So we will configure fully-meshed GRE tunnels over MPLS for support multicasting.

During normal time, Sites A, C, D will use Site A's mulitcast source for the thinking of bandwidth to site A from remote sites, B for sure will use B's source because it is local. But in case remote sites lose the connection to site A (or site A's source is not available---I will not consider application backup because the application itself can not support redundancy), those sites will be able to use site B's source. So I would setup two RPs which are site A and Site B's routers; and Site C, D will primarily use Site A's RP for multicast. In case lost connection, site C,D will use Site B's RP. ---- Site B's RP is a backup RP for site A. So I would use anycast RP but site A's RP is a primary one.

Please provide suggestion for this requirement if anycast can not use or how to use anycast in this case.

Thanks,

JJ

[Harold Ritter]

It doesn't matter which RP you select since the traffic will shift from the shared tree to the source tree after the DR connected to the receiver starts receiving trafic from the source, unless you force the traffic to stay on the shared tree.

You could set the IGP metric so that remote sites would always prefer the Anycast RP in site A and then force the traffic to remain on the shared tree. This way the remote sites would always go to the RP in site A and therefore receive multicast traffic from site A. If site A goes down the shared tree would shift from the RP in site A to the one in site B.

To force the multicast traffic to remain on the shared tree, you need to use the following command:

ip pim spt-threshold infinity

For more information on this command, please refer to the Cisco IOS documentation:

http://www.cisco.com/univercd/cc/td/doc/product/software/ios124/124tcr/timc_r/mlt_i2ht.htm#wp1070451

Hope this helps,

[jjia]

Thanks,

But I have another question, since we configured GRE tunnels for multicast traffic as per http://www.cisco.com/en/US/partner/products/ps6350/products_configuration_guide_chapter09186a00805a4e91.html,

on remote site routers we need define ip mroutes for RPF check for RP address and source address.

In our example, Anycast RP address 172.21.1.2/32 is configured on Router A, and RouterB, source address 10.1.2.22/32 on site A and source address 10.2.9.22/32 on site B are installed, and both of the routers A/B have tunnels to site C routerC. So on Router C it has tunnel0 to RouterA, and tunnel2 to RouterB.

My question is how I can input the same ip mroutes with different interfaces w/ different distance: (in C's unicast routing table,it has all /32 host routes existed for example)

ip mroute 172.21.1.2 255.255.255.255 tunnel0

ip mroute 172.21.1.2 255.255.255.255 tunnel2 120 (for backup)

ip mroute 10.1.2.22 255.255.255.255 tunnel0

ip mroute 10.1.2.22 255.255.255.255 tunnel2 120 (for backup)

ip mroute 10.2.9.22 255.255.255.255 tunnel2

ip mroute 10.2.9.22 255.255.255.255 tunnel0 120 (for backup)

But whenever I input the backup mroute, the first primary mroute has been overidden. That means I can only input one primary mroute; but if the tunnel0 is down, how to do RPF check to tunnel2?

Although I can configure the following:

ip mroute 172.21.1.2 255.255.255.255 tunnel0

ip mroute 172.21.1.0 255.255.255.0 tunnel2 120 (for backup)

ip mroute 10.1.2.22 255.255.255.255 tunnel0

ip mroute 10.1.2.0 255.255.255.0 tunnel2 120 (for backup)

ip mroute 10.2.9.22 255.255.255.255 tunnel2

ip mroute 10.2.9.0 255.255.255.0 tunnel0 120 (for backup)

not sure if this configuration will work if tunnel0 down?

or any other solutions?

thanks,

JJ

[Harold Ritter]

JJ,

I am not sure you need 2 tunnels to start with. If the primary IP path goes down, the GRE tunnel should converge through the secondary path.

Hope this helps,

[Harold Ritter]

JJ,

Forget my last reply. I had forgottten about the original design.

The technique you use with the more specific mroute for the primary and the less specific for the backup should work as long as the primary tunnel goes down.

Hope this helps,