Multicast RPF Configuration Guidelines

You use multicast reverse path forwarding (RPF) checks to prevent multicast routing loops. Routing loops are particularly debilitating in multicast applications because packets are replicated with each pass around the routing loop.

In general, a router is to forward a multicast packet only if it arrives on the interface closest (as defined by a unicast routing protocol) to the origin of the packet, whether source host or rendezvous point (RP). In other words, if a unicast packet would be sent to the “destination” (the reverse path) on the interface that the multicast packet arrived on, the packet passes the RPF check and is processed. Multicast (or unicast) packets that fail the RPF check are not forwarded (this is the default behavior). For an overview of how a Juniper Networks router implements RPF checks with tables, see Multicast RPF Checks.

However, there are network router configurations where multicast packets that fail the RPF check need to be forwarded. For example, when point-to-multipoint label-switched paths (LSPs) are used for distributing multicast traffic to PIM “islands” downstream from the egress router, the interface on which the multicast traffic arrives is not always the RPF interface. This is because LSPs do not follow the normal next-hop rules of independent packet routing. For information on LSPs, see the Junos MPLS Applications Configuration Guide.

In cases such as these, you can configure policies on the PE router to decide which multicast groups and sources are exempt from the default RPF check.

For more information about policies, see the Junos Policy Framework Configuration Guide.