- Create a new “dev” cluster - Recreate staging and prod clusters to match - Ensure that we can easily control deployment to prod/staging/dev and document this - Have simple controls to restart things as needed