Not sure of the purpose of going with repmgr. We have been using repmgr and my personal resentment with it is that every time there is a network issue, we end up with two primary and manual intervention is needed to overcome that.
That's to be expected with only two nodes. You can "fix" that by adding a third node for quorum. (It doesn't even need Postgresql on it; just has to "be there", and running heartbeat software.)
