Fix and improve dockerfile

.dockerignore

@@ -2,6 +2,7 @@
 
 !src
 !notebooks
+notebooks/.ipynb_checkpoints
 !setup.py
 !scripts
 !build_scripts

Dockerfile

@@ -1,5 +1,5 @@
 #-------------- Base Image -------------------
-FROM jupyter/minimal-notebook:python-3.11 as BASE
+FROM jupyter/minimal-notebook:python-3.10 as BASE
 
 ARG CODE_DIR=/tmp/code
 ARG POETRY_VERSION=1.8.2
@@ -19,24 +19,31 @@ ENV PATH="${POETRY_HOME}/bin:$PATH"
 
 USER root
 
-RUN curl -sSL https://install.python-poetry.org | python -
+# make is needed to install matplotlib 3.5.0 for python 3.11
+RUN apt-get update \
+    && apt-get install -y --no-install-recommends curl make \
+    && rm -rf /var/lib/apt/lists/*
+
+RUN curl -sSL https://install.python-poetry.org | python - \
+    && poetry self add poetry-plugin-export
 
 USER ${NB_UID}
 
 WORKDIR $CODE_DIR
 
-COPY --chown=${NB_UID}:${NB_GID} poetry.lock pyproject.toml .
+COPY --chown=${NB_UID}:${NB_GID} poetry.lock pyproject.toml ./
 
-RUN poetry install --no-interaction --no-ansi --no-root --without=dev
+RUN poetry export --no-interaction --no-ansi --without=dev -f requirements.txt --output requirements.txt \
+    && pip install --no-cache-dir --requirement requirements.txt
 
 COPY --chown=${NB_UID}:${NB_GID} src/ src/
 COPY --chown=${NB_UID}:${NB_GID} README.md .
 
+# Build code
 RUN poetry build
 
-
 #-------------- Main Image -------------------
-FROM jupyter/minimal-notebook:python-3.11 as MAIN
+FROM jupyter/minimal-notebook:python-3.10 as MAIN
 
 ARG CODE_DIR=/tmp/code
 
@@ -46,7 +53,7 @@ ENV DEBIAN_FRONTEND=noninteractive\
     PYTHONFAULTHANDLER=1 \
     CODE_DIR=$CODE_DIR
 
-ENV PATH="${CODE_DIR}/.venv/bin:$PATH"
+# ENV PATH="${CODE_DIR}/.venv/bin:$PATH"
 
 USER root
 
@@ -61,34 +68,33 @@ USER ${NB_UID}
 
 WORKDIR ${CODE_DIR}
 
-# Copy virtual environment from base image
-COPY --from=BASE ${CODE_DIR}/.venv ${CODE_DIR}/.venv
+# Copy conda directory from base image
+COPY --from=BASE "${CONDA_DIR}" "${CONDA_DIR}"
 # Copy built package from base image
 COPY --from=BASE ${CODE_DIR}/dist ${CODE_DIR}/dist
+# Install the built package
+RUN pip install --no-cache-dir dist/*.whl
+# Install kernel
+RUN ipython kernel install --name "tfl-training-ml-control" --user
 
-# Start of HACK: the home directory is overwritten by a mount when a jhub server is started off this image
+############## Start of HACK
+# the home directory is overwritten by a mount when a jhub server is started off this image
 # Thus, we create a jovyan-owned directory to which we copy the code and then move it to the home dir as part
 # of the entrypoint
-COPY --chown=${NB_UID}:${NB_GID} entrypoint.sh $CODE_DIR
+COPY --chown=${NB_UID}:${NB_GID} entrypoint.sh /usr/local/bin/
 
-RUN chmod +x "${CODE_DIR}/"entrypoint.sh
+RUN chmod +x "/usr/local/bin/entrypoint.sh"
 # Unfortunately, we cannot use ${CODE_DIR} in the ENTRYPOINT directive, so we have to hardcode it
 # Keep in sync with the value of CODE_DIR above
-ENTRYPOINT ["/tmp/code/entrypoint.sh"]
-
-# End of HACK
+ENTRYPOINT ["/usr/local/bin/entrypoint.sh"]
 
-WORKDIR "${HOME}"
+############## End of HACK
 
-COPY --chown=${NB_UID}:${NB_GID} . $CODE_DIR
-
-# Move to the code dir to install dependencies as the CODE_DIR contains the
-# complete code base, including the poetry.lock file 
-WORKDIR $CODE_DIR
-
-RUN pip install --no-cache-dir dist/*.whl
-
-RUN ipython kernel install --name "tfl-training-ml-control" --user
+WORKDIR "${CODE_DIR}"
 
+# Copy notebooks
+COPY --chown=${NB_UID}:${NB_GID} . .
+# Trust all notebooks
 RUN find notebooks -name '*.ipynb' -exec jupyter trust {} \;
 
+WORKDIR "${HOME}"

README.md

@@ -40,7 +40,7 @@ There are multiple ways of viewing/executing the content.
    You can then start the container e.g., with
 
    ```shell
-   docker run -it -p 8888:8888 tfl-training-ml-control:local jupyter notebook --ip=0.0.0.0
+   docker run -it -p 8888:8888 tfl-training-ml-control:local jupyter lab --ip=0.0.0.0
    ```
 
 4. Finally, for creating source code documentation, you can run
@@ -66,7 +66,7 @@ There are multiple ways of viewing/executing the content.
    ```shell
    docker run -it --rm --privileged --net=host \
       --env DISPLAY --volume /tmp/.X11-unix:/tmp/.X11-unix \
-      tfl-training-ml-control:local jupyter notebook --ip=0.0.0.0
+      tfl-training-ml-control:local jupyter lab --ip=0.0.0.0
    ```
 
 > **Note** There is some non-trivial logic in the entrypoint that may collide

entrypoint.sh

@@ -8,11 +8,11 @@ if [ ! -d  "${ROOT_DIR}" ]; then
   echo "Code not found in ${ROOT_DIR}, copying it during entrypoint. With jupyterhub this should happen only once"
   mkdir "${ROOT_DIR}"
   cp -rf "${CODE_DIR}"/* "${ROOT_DIR}/"
+  find notebooks -name '*.ipynb' -exec jupyter trust {} \;
 fi
 
 cd "${ROOT_DIR}" || exit
 
-
-# original entrypoint, see https://github.com/jupyter/docker-stacks/blob/master/base-notebook/Dockerfile#L150
+# original entrypoint, see https://github.com/jupyter/docker-stacks/blob/main/images/docker-stacks-foundation/Dockerfile#L131
 # need -s option for tini to work properly when started not as PID 1
-tini -s -g -- "$@"
+tini -s -g -- start.sh "$@"