I was taking a look at this doc:
but couple of doubts arise with the config example (Figure 2)
Do we need to enable pim (ip pim spar) between RP1 and RP2?
Do we need to enable ip pim in loopback0 or loopback1?
1) Do we need to enable pim (ip pim spar) between RP1 and RP2?
yes: good note
a multicast capable topology is needed: think of a source near RP1 that needs to send traffic to a receiver near RP2
2) Do we need to enable ip pim in loopback0 or loopback1?
not in this example that use anycast RP but with static mapping
you need to enable ip pim on the loopback of AutoRP mapping agent if you use AutoRP.
To be a candidate RP is just enough to add
ip pim rp-address
(to be noted in the past was said not to put this command on the RP itself, newer documentation puts it on the RP node(s))
To be noted that you can implement AutoRP with:
static mapping as in this example
with AutoRP that is most classical example
with BSR bootstrap protocol the more modern dynamic group to RP mapping protocol ( I've testes this successfully).
both the loopbacks (the unique and the shared one) have to be advertised in the unicast routing protocol in use.
Hope to help