In a VPC scenario, to ensure the fast multicast convergence in case of VPC fail-over, the pre-build SPT feature can be used. We can pre-build the SPT for all known (S,G) in the routing table by triggering PIM joins upstream. Use the 'ip pim pre-build-spt' command to pre-build the SPT even in the absence of any receivers.
By default, PIM (S,G) joins are triggered upstream only if the OIF-list for the (S,G) is not empty. However, in certain scenarios, for example, on the virtual port-channel (vPC) non-forwarding router, it is useful to pre-build the SPTs and maintain the (S,G) states even when the system is not forwarding on these routes. This feature makes both vPC peer switches to join the SPT, even though only one vPC peer switch actually routes the multicast traffic into the vPC domain. As both the switches part of vPC have (S,G) state, the multicast convergence is greatly improved in the case of vPC fail-over.
The disadvantage with Pre-build SPT is that it results in the multicast traffic passing over two parallel paths from the source to the vPC switch pair, consuming bandwidth on both paths. Additionally, when both vPC peer switches join the SPT, one or more upstream switches in the network may be required to perform additional multicast replications to deliver the traffic on both parallel paths toward the receivers in the vPC domain.
The general practice is that if multicast convergence can be improved without sacrificing performance, then it would be recommended to use this the "pre-build" feature. The key would be to make sure that all of the requirements to enable this are met. The "requirements" meaning understanding that whatever the device that is placed upstream to vPC peers, has the ability to double its replication as that device will be performing 2x the replications as compared single replication prior to enabling pre-build on the downstream vPC peers.
The upstream router will have one OIF (pointing to one 7k in vPC)for its multicast entries, and when we enable the "pre-build" on the 7k's, then the upstream router will now create a second OIF, therefore it
(a) increases the amount of replications it has to do, and (b) now utilizes bandwidth on both parallel links, in which one of them will be dropped at the non-forwarding 7k (due to no OIF)