As described in the blog post about Anycast RP, we can take advantage of MSDP to have a load-balancing & Failover mechanism between RP, but still, we need to configure the RP address on each and every device taking part of the multicast infrastructure.
To automate this repetitive (and boring) task, Cisco has developed Auto-RP, a way to automatically elect a RP. Although we can use Auto-RP “on-top” of Anycast RP, both can be configured independently.
Auto-RP is composed of two main components: Candidate RP & RP Mapping Agent
The Candidate RP announces itself as RP via RP Announce packets to the 224.0.1.39 multicast group.
The RP Mapping Agent joins the 224.0.1.39 multicast group to receive RP Announce in order to have a group-to-RP mapping cache, it will then elect the candidate RP (based on the highest IP address) as RP for the group. The RP Mapping Agent finally, advertise the RP to the rest of the network via RP Discovery Messages to the 224.0.1.40 multicast address.
As we see, Auto-RP relies on multicast to work, however, in a PIM Sparse-Mode multicast infrastructure, a RP is needed for source registration. To solve this causality dilemma, we have three options:
- Use PIM Sparse-Dense mode
- “Flood & Prune” method for groups without RP assigned
- Static RP assignment for 224.0.1.39 & 224.0.1.40 groups
- In contradiction with the principle of automation
- Use Auto-RP listener command
Let’s go with the third option which suits best our needs, the configuration is really straightforward:
- On the Candidate RP
ip pim send-rp-announce interface scope ttl-of-rp-announce
- On the RP Mapping Agent
ip pim send-rp-discovery interface scope ttl-of-rp-discovery
- On All routers within the multicast infrastructure
ip pim autorp listener
Notes:
- Scope (TTL) must be defined to limit the range of propagation of RP-Announce & RP-Discovery, creating “limited area” with elected-RP on each;
- Auto-RP and BSR shouldn’t coexist on the same multicast infrastructure;
- It’s possible to have multiple RP Mapping Agent for redundancy;
- Specify loopback interface to send RP-announce/discovery, this avoid to have duplicate entry for the same Candidate RP at the RP Mapping Agent level;
- If deployed with Anycast-RP, do not use the Anycast-RP loopback interface, prefer the MSDP peering loopback interface;
- Candidate RP can send RP-Announce for a specific group. By default, they are sent for the whole 224.0.0.0/4 group.
Troubleshoot:
- show ip pim rp mapping on candidate RP should display the RP by multicast group with the indication “elected via Auto-RP”
- show ip pim rp mapping on RP Mapping agent should display the RP Candidates and the elected RP
- show ip pim autorp display if Auto-RP is enable and provide some statistics
- debug ip pim auto-rp is useful to monitor the activity between Candidate RP and RP Mapping Agent
Sources:
Daniel Lardeux’s Blog
Cisco – Rendezvous Point Engineering