Parabricks jupyter (#229)

* initial commits for the nemo container custom app

* adding Dockerfile using nvcr.io/nvidia/nemo:24.09

* putting the image directly in docker-compose.yaml

* commented out the sudo command at the end

* fixed syntax error with network spacing, set user to root

* fixed more indentation errors

* indentation fix not persisting?

* removed initialize command, hardcoded /workspace as the post-startup.sh script, added java and google-cloud-cli as features to install

* added the jupyter command back in

* adding files for parabricks container, only differs from nemo in the image itself right now

* modifying parabricks container to run command: tail -f /dev/null

* adding a workbench-jupyter-parabricks directory

* adding restart: always to the parabricks app

* replacing wb with /usr/bin/wb to avoid conflicts with the weights&biases executable

* Removing parabricks standalone

* alias the weightsbiases in the postCreateCommand in the devcontainer for nemo_jupyter

* removing erroneous wb replacement

* replacing more instances of wb with /usr/bin/wb

* replacing more instances of wb

* replaced /usr/bin/wb with

* added  mappings back into .devcontainer.json

* updating devcontainer-template id and description

* adding quotes around all calls to {WORKBENCH_INSTALL_PATH}

* updated some descriptions and IDs

* using a Dockerfile that now starts from the parabricks container base image and installs Jupyter

* updating postCreateCommand to match the nemo custom app

* Adding mise and java features

* updated Dockerfile to set the user as jupyter and set the home directory to /home/jupyter, use this new image in docker-compose.yaml

* tried an approach that doesn't quite get it right, will revamp

* updating Dockerfile to use jupyter user and set home to working directory, removed --allow-root from docker-compose.yaml

* create group first, update docker-compose with the new image tag

* using build context

* adding local bin to path

* removing some of the extraneous commands from the Dockerfile, switching back to the older java feature in the devcontainer

* switching to the old java digest

* use jupyter and /home/jupyter for the user and workspace, add sudo-passwordless.sh and call it in the devcontainer postCreateCommand

* removing unused version of mise-java

* updated Dockerfile per PR comments, adding sudo for all users and changing the shell to bash

* removing unused mise, updated devcontainer template id

* added missing mkdir command

* using SHELL instruction to try and set bash as the default shell, cleaned up subsequent RUN command

* still trying to update the Dockerfile to get the default shell within Jupyter to be bash

* Reconfiguring order

* removed sudo-passwordless.sh and corresponding call in postCreateCommand, adding aws-cli in as a feature

* updating command to attempt removal of the bash -e not found error

* adding parabricks to the test-pr detect-changes job

* trying to just use ENV SHELL to set the shell terminal

* changing escape sequences command

---------

Co-authored-by: Yu Hu <[email protected]>

Author: PeterSu92

.github/workflows/test-pr.yaml

@@ -54,6 +54,10 @@ jobs:
               - './**/nemo_jupyter/**'
               - 'startupscript/**'
               - 'test/**'
+            workbench-jupyter-parabricks:
+              - './**/workbench-jupyter-parabricks/**'
+              - 'startupscript/**'
+              - 'test/**'
 
   test:
     needs: [detect-changes]

src/workbench-jupyter-parabricks/.devcontainer.json

@@ -0,0 +1,58 @@
+{
+  "name": "Workbench JupyterLab with docker support devcontainer template plus NVIDIA Parabricks",
+  "dockerComposeFile": "docker-compose.yaml",
+  "service": "app",
+  "shutdownAction": "none",
+  "workspaceFolder": "/workspace",
+  // Get the host's docker group ID and propagate it into the .env file, which
+  // allows it to be used within docker-compose.yaml.
+  "initializeCommand": "DOCKER_GID=`getent group docker | cut -d: -f3` && echo \"DOCKER_GID=${DOCKER_GID}\" > .env",
+  "postCreateCommand": "apt update && apt install -y sudo && ./startupscript/post-startup.sh jupyter /home/jupyter ${templateOption:cloud} ${templateOption:login} && printf 'alias weightsbiases=\"/usr/local/bin/wb\"\\nalias wb=\"/usr/bin/wb\"\\n' >> /home/jupyter/.bashrc",
+  // re-mount bucket files on container start up
+  "postStartCommand": [
+    "./startupscript/remount-on-restart.sh",
+    "jupyter",
+    "/home/jupyter",
+    "${templateOption:cloud}",
+    "${templateOption:login}"
+  ],
+  "features": {
+    "ghcr.io/devcontainers/features/java@sha256:df67d6ff6e9cdd858207ae9e92a99ddb88384b789f79eecd6f873216e951d286": {
+          "version": "17"
+        },
+    "ghcr.io/dhoeric/features/google-cloud-cli@sha256:fa5d894718825c5ad8009ac8f2c9f0cea3d1661eb108a9d465cba9f3fc48965f": {},
+    "ghcr.io/devcontainers/features/aws-cli@sha256:bbc9fd513c22e331953126c75ad7b2ed1f9044f1cd5890b7073b634810459b18":{}
+  },
+  "remoteUser": "root",
+  "customizations": {
+    "workbench": {
+      "opens": {
+        "extensions": [
+          // Source
+          ".ipynb",
+          ".R",
+          ".py",
+          // Documents
+          ".md",
+          ".html",
+          ".latex",
+          ".pdf",
+          // Images
+          ".bmp",
+          ".gif",
+          ".jpeg",
+          ".jpg",
+          ".png",
+          ".svg",
+          // Data
+          ".csv",
+          ".tsv",
+          ".json",
+          ".vl"
+        ],
+        "fileUrlSuffix": "/lab/tree/{path}",
+        "folderUrlSuffix": "/lab/tree/{path}"
+      }
+    }
+  }
+}

src/workbench-jupyter-parabricks/Dockerfile

@@ -0,0 +1,48 @@
+# Use the official NVIDIA Clara Parabricks container as the base image
+FROM nvcr.io/nvidia/clara/clara-parabricks:4.5.1-1
+
+#  Set the default shell for subsequent commands to bash.
+SHELL ["/bin/bash", "-o", "pipefail", "-c"]
+
+ARG NB_USER=jupyter
+ARG NB_UID=1010
+ARG NB_GID=1000
+ARG WORKDIR=/home/${NB_USER}
+
+# Switch to root user to install packages and create new user
+USER root
+
+# This folder caused java installation issues, removing it resolves this
+RUN rm -rf /var/lib/apt/lists/*
+
+RUN \
+    mkdir -p /etc/sudoers.d/ \
+    # Create the user and set their shell directly in one step.
+    # This avoids potential errors with 'groupadd' and makes 'chsh' redundant.
+    && groupadd --gid ${NB_GID} ${NB_USER} && useradd --uid ${NB_UID} --gid ${NB_GID} --create-home --home-dir ${WORKDIR} --shell /bin/bash ${NB_USER} \
+    # Grant password-less sudo access to the new user
+    && echo "${NB_USER} ALL=(ALL) NOPASSWD:ALL" > /etc/sudoers.d/${NB_USER} \
+    && chmod 0440 /etc/sudoers.d/${NB_USER}
+
+# Fix ownership for common dirs
+RUN mkdir -p /workspace \
+ && chown -R ${NB_UID}:${NB_GID} ${WORKDIR} /workspace /tmp \
+ && chown -R ${NB_UID}:${NB_GID} /opt/conda || true
+
+
+
+# Environment and working directory
+ENV HOME=${WORKDIR}
+WORKDIR ${WORKDIR}
+
+ENV PATH="${WORKDIR}/.local/bin:${PATH}"
+# Switch to non-root user
+USER ${NB_USER}
+ENV SHELL=/bin/bash
+
+# Create a Jupyter config file to force the terminal to use bash
+# RUN mkdir -p ~/.jupyter/ && \
+#     echo "c.ServerApp.terminado_settings = {'shell_command': ['/bin/bash']}" > ~/.jupyter/jupyter_server_config.py
+
+# Install Jupyter Lab using pip
+RUN pip3 install jupyterlab

src/workbench-jupyter-parabricks/devcontainer-template.json

@@ -0,0 +1,23 @@
+{
+  "id": "Jupyterlab on Parabricks",
+  "description": "A template used to serve the Workbench JupyterLab container image plus NVIDIA Parabricks",
+  "version": "0.0.1",
+  "name": "NVIDIA Parabricks container with Jupyterlab installed",
+  "documentationURL": "https://github.com/verily-src/workbench-app-devcontainers/tree/master/src/custom-workbench-jupyter-template",
+  "licenseURL": "https://github.com/verily-src/workbench-app-devcontainers/blob/master/LICENSE",
+  "options": {
+    "cloud": {
+      "type": "string",
+      "description": "VM cloud environment",
+      "proposals": ["gcp", "aws"],
+      "default": "gcp"
+    },
+    "login": {
+      "type": "string",
+      "description": "Whether to log in to workbench CLI",
+      "proposals": ["true", "false"],
+      "default": "false"
+    }
+  },
+  "platforms": ["Any"]
+}

src/workbench-jupyter-parabricks/docker-compose.yaml

@@ -0,0 +1,24 @@
+version: "2.4"
+services:
+  app:
+    container_name: "application-server"
+    build:
+      context: .
+    user: jupyter
+    restart: always
+    volumes:
+      - .:/workspace:cached
+    ports:
+      - "8888:8888"
+    networks:
+      - app-network
+    cap_add:
+      - SYS_ADMIN
+    devices:
+      - /dev/fuse
+    security_opt:
+      - apparmor:unconfined
+    command: ["jupyter", "lab", "--ip=0.0.0.0", "--port=8888", "--no-browser", "--LabApp.token=''"]
+networks:
+  app-network:
+    external: true