There are basically three types of multicast routing protocols: broadcast and prune, membership advertisement and rendezvous-based whose examples can be DVMRP, MOSPF and PIM-SM respectively. In rendezvous-based protocol, each multicast group address is mapped to a address called ?core? or ?rendezvous-point? (RP). Members willing to receive data sent to the group .rst join the RP. When a RP fails, a new RP should be selected and the multicast group should be mapped to it. In this paper, unlike bootstrap mechanism which redistributes RP-set should a RP fails, we propose a mechanism to construct a RP table at each PIM-SM router and to distribute a pair of (RP_primary, RP_secondaries) which is a column of the RP table. In this mechanism initially a multicast group addresses are mapped to RP_primaries. When a RP_primary fails corresponding RP_secondary takes over the function of the failed RP_primary (i.e. becomes RP_primary), selects a new RP_secondary from the table, updates the table with the new (RP_primary, RP_secondary) and distributes the new (RP_primary, RP_secondary) to all routers in the domain. When the RP_secondary) fails the RP_primary selects a new RP_secondary from the table, updates the table from the new (RP_primary, RP_secondaries) and distributes the new (RP_primary, RP_secondaries) to all routers in the domain. When the RP_secondaries fails the RP_primary selects a new RP_secondaries from the table, updates the table and distributes the new (RP_primary, RP_secondaries) pair to all routers. Each router in the domain after receiving the (RP_primary, RP_secondaries) updates its table accordingly. RP_primary, RP_secondaries exchange live message periodically to check each others liveliness.