Strategy to update OSM maps while avoiding the server downtime

I’m looking to avoid the downtime of a GraphHopper server when updating OSM maps (by updating I mean different regions’ maps have to be used). If my understanding is correct, maps can only be built during the compilation time. Plus, if multiple regions have to be stitched together, the downtime could be significant.

When hosting a GraphHopper server in an AWS EC2 instance, what is the strategy to avoid the downtime when the maps have to be updated periodically?

You can run the map imports on a separate machine and copy over the resulting files.

When you say resulting files, do you mean everything within the graphs directory?

Yes. The server will start rather quickly if the files are already present.