Building custom images to install R and Python packages

You can build custom images based on the SPSS Modeler runtime images available in IBM Watson Studio.

Before you begin

You must have a Cloud Pak for Data cluster administrator role to create and register a custom image.

Note: If you use custom images, it is your responsibility to update your custom images with all the latest updates that are made to the available runtime images in Watson Studio. You are responsible for installing new fix packs, security updates, or any other updates. When new versions of Cloud Pak for Data are released, rebuilding all of your custom images is the best way to keep them updated.

About this task

You can use a custom image to optimize the standard configuration of an SPSS Modeler runtime. You can also use custom images in air-gapped environments that cannot expose any operations to the internet.

Procedure

Prepare to build a new image.
1. Download the configuration file for the SPSS Modeler runtime image that you want to customize. See Downloading the configuration file.
2. Download the image in the configuration. See Downloading the image.
3. Get the registry URL from the IBM Entitled registry or a private Container Registry.
  The SPSS Modeler runtime images are stored in a Docker image registry. In Cloud Pak for Data, you can use only an external registry outside of the Cloud Pak for Data Red Hat OpenShift cluster. To use that registry, you need the URL to the external registry that was used during the installation of Cloud Pak for Data. Use the same URL for all commands and in all the scripts that you run.
4. Download the build-image-for-python-r-packages.zip that contains the Dockerfile and scripts for installing R and Python packages.
  The archive contains the following files:
  - install-packages.R is the main script to install R packages.
  - install-Python-packages.sh is the main script to install Python packages.
  - Dockerfile is the main entry to build the custom image.
Create the custom runtime image with the files that you downloaded, and add your R or Python packages.
1. Add customizations and build a new image. See Creating a custom image.
2. In Dockerfile from the archive, update <spss-modeler-base-image> with the tag of the actual spss-modeler base image that you pulled from the IBM Entitled registry or a private Container Registry.
  Example of the Dockerfile:
```
FROM <spss-modeler-base-image> as custom-runtime-for-python-r-packages

USER modeler-user:modeler-user

COPY --chown=modeler-user install-packages.R /tmp/install/
COPY --chown=modeler-user install-Python-packages.sh /tmp/install/

RUN umask 002 \
 && /opt/r/lib/R/bin/Rscript /tmp/install/install-packages.R \
 && bash /tmp/install/install-Python-packages.sh \
 && rm -rf /tmp/install
```
3. If you are installing R packages, add all your package names to the list in install-packages.R. Every package name must be in double quotation marks.
```
install.packages(c(
"pkg-name1",
"pkg-name2",
 ...),
repos='http://cran.rstudio.com')
```
4. If you are installing Python packages, add all your package names after the pip install command in install-Python-packages.sh.
```
set -e -o pipefail
MODELER_HOME=/opt/IBM/SPSS/ModelerServer/Cloud/
LD_LIBRARY_PATH=$MODELER_HOME/python/lib $MODELER_HOME/python_venv/bin/pip install \
  pkg-name1 \
  pkg-name2 \
  ...
```
5. If you want to install only R packages or only Python packages, comment out the line in Dockerfile for the installation that you want to skip.
  By default, Dockerfile invokes both install-packages.R and install-Python-packages.sh to install the R and Python packages together.
  - To install only R packages, comment out && bash /tmp/install/install-Python-packages.sh \.
  - To install only Python packages, comment out && /opt/r/lib/R/bin/Rscript /tmp/install/install-packages.R \.
Push the image to the container server to register it. See Registering the custom image.
Change and upload the configuration file to use the new custom image. See Uploading the changed configuration.