Docker for Jupyter Notebook tutorials

Apache Druid provides a custom Jupyter container that contains the prerequisites for all Jupyter-based Druid tutorials, as well as all of the tutorials themselves. You can run the Jupyter container, as well as containers for Druid and Apache Kafka, using the Docker Compose file provided in the Druid GitHub repository.

You can run the following combination of applications:

Prerequisites

Jupyter in Docker requires that you have Docker and Docker Compose. We recommend installing these through Docker Desktop.

For ARM-based devices, see Tutorial setup for ARM-based devices.

Launch the Docker containers

You run Docker Compose to launch Jupyter and optionally Druid or Kafka. Docker Compose references the configuration in docker-compose.yaml. Running Druid in Docker also requires the environment file, which sets the configuration properties for the Druid services. To get started, download both docker-compose.yaml and environment from tutorial-jupyter-docker.zip.

Alternatively, you can clone the Apache Druid repo and access the files in druid/examples/quickstart/jupyter-notebooks/docker-jupyter.

Start only the Jupyter container

If you already have Druid running locally or on another machine, you can run the Docker containers for Jupyter only. In the same directory as docker-compose.yaml, start the application:

  1. docker compose --profile jupyter up -d

The Docker Compose file assigns 8889 for the Jupyter port. You can override the port number by setting the JUPYTER_PORT environment variable before starting the Docker application.

If Druid is running local to the same machine as Jupyter, open the tutorial and set the host variable to host.docker.internal before starting. For example:

  1. host = "host.docker.internal"

Start Jupyter and Druid

Running Druid in Docker requires the environment file as well as an environment variable named DRUID_VERSION, which determines the version of Druid to use. The Druid version references the Docker tag to pull from the Apache Druid Docker Hub.

In the same directory as docker-compose.yaml and environment, start the application:

  1. DRUID_VERSION=28.0.0 docker compose --profile druid-jupyter up -d

Start Jupyter, Druid, and Kafka

Running Druid in Docker requires the environment file as well as the DRUID_VERSION environment variable.

In the same directory as docker-compose.yaml and environment, start the application:

  1. DRUID_VERSION=28.0.0 docker compose --profile all-services up -d

Start Kafka and Jupyter

If you already have Druid running externally, such as an existing cluster or a dedicated infrastructure for Druid, you can run the Docker containers for Kafka and Jupyter only.

In the same directory as docker-compose.yaml and environment, start the application:

  1. DRUID_VERSION=28.0.0 docker compose --profile kafka-jupyter up -d

If you have an external Druid instance running on a different machine than the one hosting the Docker Compose environment, change the host variable in the notebook tutorial to the hostname or address of the machine where Druid is running.

If Druid is running local to the same machine as Jupyter, open the tutorial and set the host variable to host.docker.internal before starting. For example:

  1. host = "host.docker.internal"

To enable Druid to ingest data from Kafka within the Docker Compose environment, update the bootstrap.servers property in the Kafka ingestion spec to localhost:9094 before ingesting. For reference, see Consumer properties.

Update image from Docker Hub

If you already have a local cache of the Jupyter image, you can update the image before running the application using the following command:

  1. docker compose pull jupyter

Use locally built image

The default Docker Compose file pulls the custom Jupyter Notebook image from a third party Docker Hub. If you prefer to build the image locally from the official source, do the following:

  1. Clone the Apache Druid repository.
  2. Navigate to examples/quickstart/jupyter-notebooks/docker-jupyter.
  3. Start the services using -f docker-compose-local.yaml in the docker compose command. For example:
  1. DRUID_VERSION=28.0.0 docker compose --profile all-services -f docker-compose-local.yaml up -d

Access Jupyter-based tutorials

The following steps show you how to access the Jupyter notebook tutorials from the Docker container. At startup, Docker creates and mounts a volume to persist data from the container to your local machine. This way you can save your work completed within the Docker container.

  1. Navigate to the notebooks at http://localhost:8889.

    Docker for tutorials - 图1info

    If you set JUPYTER_PORT to another port number, replace 8889 with the value of the Jupyter port.

  2. Select a tutorial. If you don’t plan to save your changes, you can use the notebook directly as is. Otherwise, continue to the next step.

  3. Optional: To save a local copy of your tutorial work, select File > Save as… from the navigation menu. Then enter work/<notebook name>.ipynb. If the notebook still displays as read only, you may need to refresh the page in your browser. Access the saved files in the notebooks folder in your local working directory.

View the Druid web console

To access the Druid web console in Docker, go to http://localhost:8888/unified-console.html. Use the web console to view datasources and ingestion tasks that you create in the tutorials.

Stop Docker containers

Shut down the Docker application using the following command:

  1. docker compose down -v

Tutorial setup without using Docker

To use the Jupyter Notebook-based tutorials without using Docker, do the following:

  1. Clone the Apache Druid repo, or download the tutorials as well as the Python client for Druid.

  2. Install the prerequisite Python packages with the following commands:

    1. # Install requests
    2. pip install requests
    1. # Install JupyterLab
    2. pip install jupyterlab
    4. # Install Jupyter Notebook
    5. pip install notebook

    Individual notebooks may list additional packages you need to install to complete the tutorial.

  3. In your Druid source repo, install druidapi with the following commands:

    1. cd examples/quickstart/jupyter-notebooks/druidapi
    2. pip install .

  4. Start Jupyter, in the same directory as the tutorials, using either JupyterLab or Jupyter Notebook:

    1. # Start JupyterLab on port 3001
    2. jupyter lab --port 3001
    4. # Start Jupyter Notebook on port 3001
    5. jupyter notebook --port 3001

  5. Start Druid. You can use the Quickstart (local) instance. The tutorials assume that you are using the quickstart, so no authentication or authorization is expected unless explicitly mentioned.

    If you contribute to Druid, and work with Druid integration tests, you can use a test cluster. Assume you have an environment variable, DRUID_DEV, which identifies your Druid source repo.

    1. cd $DRUID_DEV
    2. ./it.sh build
    3. ./it.sh image
    4. ./it.sh up <category>

    Replace <category> with one of the available integration test categories. See the integration test README.md for details.

You should now be able to access and complete the tutorials.

Tutorial setup for ARM-based devices

For ARM-based devices, follow this setup to start Druid externally, while keeping Kafka and Jupyter within the Docker Compose environment:

  1. Start Druid using the start-druid script. You can follow Quickstart (local) instructions. The tutorials assume that you are using the quickstart, so no authentication or authorization is expected unless explicitly mentioned.

  2. Start either Jupyter only or Jupyter and Kafka using the following commands in the same directory as docker-compose.yaml and environment:

    1. # Start only Jupyter
    2. docker compose --profile jupyter up -d
    4. # Start Kafka and Jupyter
    5. DRUID_VERSION=28.0.0 docker compose --profile kafka-jupyter up -d

  3. If Druid is running local to the same machine as Jupyter, open the tutorial and set the host variable to host.docker.internal before starting. For example:

    1. host = "host.docker.internal"

  4. If using Kafka to handle the data stream that will be ingested into Druid and Druid is running local to the same machine, update the consumer property bootstrap.servers to localhost:9094.

Learn more

See the following topics for more information: