Custom Builds

Example for a Custom Build of ngen

Here are steps for making the RTE image using a custom version of ngen itself. These steps assume that the default version of ngen-forcing (referenced by ngen/Dockerfile) is the desired base, prior to adding the ngen layer.

1. Run the above setup steps for cloning repos and downloading data, except stop before running ./ngen_rte_build.sh.

2. Edit config.bashrc to use: NGEN_SOURCE_MODE="build_from_local"

3. Check out the desired code for ngen and its submodules. For example, to use ngen's development branch but change the version of t-route:

   ### From the ~/ngwpc/ngen repository root:
   git checkout development
   git pull
   ## Ensure the local copies of submodules are matching the specific commits currently pointed to by the ngen repository.
   ## This does not fetch the latest updates from the submodule branches tracked by .gitmodules.  To do that, you would add --remote to the command.
   git submodule update --init --recursive
   ## Check out a specific branch of t-route
   ( cd extern/t-route && git checkout any-branch && git pull )
   

4. If your ngen/Dockerfile is starting FROM ghcr.io/ngwpc/ngen-bmi-forcing:latest, then pull the latest ngen-forcing GHCR image.

   sudo docker pull ghcr.io/ngwpc/ngen-bmi-forcing:latest
   

5. Build the RTE image

   ### From the ~/ngwpc/nwm-rte repository root:
   ./ngen_rte_build.sh
   

6. Confirm in the build messages that the built was successful and that the image name built is ngen_rte_build_from_local

See Also

For (re)installing the Python packages of ngen's submodules, such as lstm, within the Dev Container, see .devcontainer/devcontainer.json.poststart.sh