I have an octree that I'm planning to use for 3D pathfinding. My plan was to break down the world such that all the leaf nodes of the tree would be of equal dimensions that would create a nice grid and then adapt A* by adding an extra dimension and just run searches on the grid. The image below shows a sample of my leaf nodes to demonstrate what I'm talking about (ignore the spheres, they are for something else).

What I'm not sure about is how do I perform A* on a tree structure. My thought was to convert the tree structure to a 3D array, which would make it really easy to perform A* on, but I'm not sure how I would go about this. I can recursively work my way down the tree and tell which nodes are the leaf nodes, but I don't know how I could necessarily work my way down in an ordered fashion such that I could index nodes into an array properly. How would I go about this? Is it even the right approach; or it there a better approach to performing A* on the tree?
Note: I am aware the space might not actually be open enough to justify full 3D pathfinding, but this is intended to be a demonstration/test of the concept, so I wanted to keep things relatively simple.