We have MongoDB environments using a 3-node replicaSet (without arbiter) in production. We have started making use of SnapCreator to take daily snapshots from the Primary node in the replicaSet.
My question is, if at any point the Primary switches between nodes would the snapshot (running from Secondary node) still be useful for recovery of the entire replicaSet? If so, what should be the steps if the snapshot has to be followed with opLog replay?
In general, what's the standard practice for using SnapCreator with MongoDB replicaSets? Snapshot from all nodes/volumes?