The Phantom RP is the RP redundancy mechanism for PIM-BiDir. The PIM Anycast RP and Anycast RP with MSDP are the mechanisms used for RP redundancy in the PIM ASM (Any Source Multicast).
In this video, I explain the PIM phantom RP with a LAB demonstration using Cisco Nexus switches.
Summary
- Phantom RP means the RP address (RPA) does not belong to any physical router’s interface but only to a reachable network.
- Phantom RP is provided using loopback interfaces configured in multiple routers, each with a different subnet mask for the RP address (none of them is the exact /32 address).
- The unicast routing longest match rule is used to choose the active RP.
- If the active RP fails, the secondary RP with the shorter mask, still advertising the RP subnet, will appear in the routing table as the best route.
- For more information: RP Redundancy with PIM Bidir – Phantom RP
I hope this article was useful. For comprehensive content, you can refer to my CCIE Data Center (v3.1) – Multicast Udemy course.
If I configure a phantom RP on each device in IPN with 4 routers (2 routers in each POD) : 2 phantom RP /30 in POD 1 and 2 phantom RP /29 in POD2 , what will be the behavior? Or you advice me to configure only 2 phantom RP (one in each pod)?
You should always have one phantom RP for any given multicast group. So, you should have one phantom RP per IPN, not per Pod. Because the tenant multicast traffic is expected to pass through the IPN devices where you need the routers to agree on the RP address for all multicast groups, right?
If you are looking for high availability, you already have it if you configure the phantom RP correctly. Once the active phantom RP goes down, the RP with the second-best route will be active. But if you want redundancy and load-balancing, you can split the multicast range between two or more active phantom RPs. So, for several multicast ranges, you need to repeat the whole configuration process with different IP addresses.
We are experiencing some line quality issue which ipn nodes not noticed link down but wait for ospf timeout. BFD came to help. Now IGP failover quickly. I am wondering would pim bidir failover quick accordingly with igp? Or do we need to implement pim bfd?
Without pim bfd, I noticed ospf went down under 1 second, but pim takes looong time.
IGP converged on new path, not sure pim bidir got blackholed or re-routed.
In short, for fast failover, would bfd for IGP good enough?
PIM bidir failover is based on the phantom RP technique, which relies on the routing table (RIB) on each PIM router in the IPN network. The faster the RIB convergence, the quicker the PIM bidir failover you will have. So, yes, IGP BFD is enough.
On the other hand, PIM BFD is all about PIM neighborship convergence, just like IGP.
Please remember that during the failover, you need a PIM-enabled route to reach the new RP. That means the PIM neighborship should already be established to consider the alternative path valid.
So, enabling PIM BFD will allow the initial (pre-issue) PIM state to be reverted faster.
Thank you Salman. 👍👌