Contributed Recipes

Users sometimes share interesting ways of using the Jupyter Docker Stacks. We encourage users to contribute these recipes to the documentation in case they prove useful to other members of the community by submitting a pull request to docs/using/ The sections below capture this knowledge.

Using sudo within a container

Password authentication is disabled for the NB_USER (e.g., jovyan). This choice was made to avoid distributing images with a weak default password that users ~might~ will forget to change before running a container on a publicly accessible host.

You can grant the within-container NB_USER passwordless sudo access by adding -e GRANT_SUDO=yes and --user root to your Docker command line or appropriate container orchestrator config.

For example:

docker run -it -e GRANT_SUDO=yes --user root jupyter/minimal-notebook

You should only enable sudo if you trust the user and/or if the container is running on an isolated host. See Docker security documentation for more information about running containers as root.

Using mamba install or pip install in a Child Docker image

Create a new Dockerfile like the one shown below.

# Start from a core stack version
FROM jupyter/datascience-notebook:33add21fab64
# Install in the default python3 environment
RUN pip install --quiet --no-cache-dir 'flake8==3.9.2' && \
    fix-permissions "${CONDA_DIR}" && \
    fix-permissions "/home/${NB_USER}"

Then build a new image.

docker build --rm -t jupyter/my-datascience-notebook .

To use a requirements.txt file, first create your requirements.txt file with the listing of packages desired. Next, create a new Dockerfile like the one shown below.

# Start from a core stack version
FROM jupyter/datascience-notebook:33add21fab64
# Install from requirements.txt file
COPY --chown=${NB_UID}:${NB_GID} requirements.txt /tmp/
RUN pip install --quiet --no-cache-dir --requirement /tmp/requirements.txt && \
    fix-permissions "${CONDA_DIR}" && \
    fix-permissions "/home/${NB_USER}"

For conda, the Dockerfile is similar:

# Start from a core stack version
FROM jupyter/datascience-notebook:33add21fab64
# Install from requirements.txt file
COPY --chown=${NB_UID}:${NB_GID} requirements.txt /tmp/
RUN mamba install --yes --file /tmp/requirements.txt && \
    mamba clean --all -f -y && \
    fix-permissions "${CONDA_DIR}" && \
    fix-permissions "/home/${NB_USER}"

Ref: docker-stacks/commit/79169618d571506304934a7b29039085e77db78c

Add a Python 2.x environment

Python 2.x was removed from all images on August 10th, 2017, starting in tag cc9feab481f7. You can add a Python 2.x environment by defining your own Dockerfile inheriting from one of the images like so:

# Choose your desired base image
FROM jupyter/scipy-notebook:latest

# Create a Python 2.x environment using conda including at least the ipython kernel
# and the kernda utility. Add any additional packages you want available for use
# in a Python 2 notebook to the first line here (e.g., pandas, matplotlib, etc.)
RUN mamba create --quiet --yes -p "${CONDA_DIR}/envs/python2" python=2.7 ipython ipykernel kernda && \
    mamba clean --all -f -y

USER root

# Create a global kernelspec in the image and modify it so that it properly activates
# the python2 conda environment.
RUN "${CONDA_DIR}/envs/python2/bin/python" -m ipykernel install && \
    "${CONDA_DIR}/envs/python2/bin/kernda" -o -y /usr/local/share/jupyter/kernels/python2/kernel.json



Add a Python 3.x environment

The default version of Python that ships with conda/ubuntu may not be the version you want. To add a conda environment with a different version and make it accessible to Jupyter, the instructions are very similar to Python 2.x but are slightly simpler (no need to switch to root):

# Choose your desired base image
FROM jupyter/minimal-notebook:latest

# name your environment and choose python 3.x version
ARG conda_env=python36
ARG py_ver=3.6

# you can add additional libraries you want mamba to install by listing them below the first line and ending with "&& \"
RUN mamba create --quiet --yes -p "${CONDA_DIR}/envs/${conda_env}" python=${py_ver} ipython ipykernel && \
    mamba clean --all -f -y

# alternatively, you can comment out the lines above and uncomment those below
# if you'd prefer to use a YAML file present in the docker build context

# COPY --chown=${NB_UID}:${NB_GID} environment.yml "/home/${NB_USER}/tmp/"
# RUN cd "/home/${NB_USER}/tmp/" && \
#     mamba env create -p "${CONDA_DIR}/envs/${conda_env}" -f environment.yml && \
#     mamba clean --all -f -y

# create Python 3.x environment and link it to jupyter
RUN "${CONDA_DIR}/envs/${conda_env}/bin/python" -m ipykernel install --user --name="${conda_env}" && \
    fix-permissions "${CONDA_DIR}" && \
    fix-permissions "/home/${NB_USER}"

# any additional pip installs can be added by uncommenting the following line
# RUN "${CONDA_DIR}/envs/${conda_env}/bin/pip" install

# prepend conda environment to path
ENV PATH "${CONDA_DIR}/envs/${conda_env}/bin:${PATH}"

# if you want this environment to be the default one, uncomment the following line:
# ENV CONDA_DEFAULT_ENV ${conda_env}

Run JupyterLab

JupyterLab is preinstalled as a notebook extension starting in tag c33a7dc0eece.

Run jupyterlab using a command such as docker run -it --rm -p 8888:8888 -e JUPYTER_ENABLE_LAB=yes jupyter/datascience-notebook

Dask JupyterLab Extension

Dask JupyterLab Extension provides a JupyterLab extension to manage Dask clusters, as well as embed Dask's dashboard plots directly into JupyterLab panes. Create the Dockerfile as:

# Start from a core stack version
FROM jupyter/scipy-notebook:latest

# Install the Dask dashboard
RUN pip install --quiet --no-cache-dir dask-labextension && \
    fix-permissions "${CONDA_DIR}" && \
    fix-permissions "/home/${NB_USER}"

# Dask Scheduler & Bokeh ports

ENTRYPOINT ["jupyter", "lab", "--ip=", "--allow-root"]

And build the image as:

docker build -t jupyter/scipy-dasklabextension:latest .

Once built, run using the command:

docker run -it --rm -p 8888:8888 -p 8787:8787 jupyter/scipy-dasklabextension:latest


Let's Encrypt a Notebook server

See the README for the simple automation here which includes steps for requesting and renewing a Let's Encrypt certificate.


Slideshows with Jupyter and RISE

RISE allows via extension to create live slideshows of your notebooks, with no conversion, adding javascript Reveal.js:

# Add Live slideshows with RISE
RUN mamba install --quiet --yes -c damianavila82 rise && \
    mamba clean --all -f -y && \
    fix-permissions "${CONDA_DIR}" && \
    fix-permissions "/home/${NB_USER}"

Credit: Paolo D. based on docker-stacks/issues/43


You need to install conda-forge's gcc for Python xgboost to work properly. Otherwise, you'll get an exception about missing GOMP_4.0.

mamba install --quiet --yes gcc && \
    mamba clean --all -f -y && \
    fix-permissions "${CONDA_DIR}" && \
    fix-permissions "/home/${NB_USER}"

pip install --quiet --no-cache-dir xgboost && \
    fix-permissions "${CONDA_DIR}" && \
    fix-permissions "/home/${NB_USER}"

# run "import xgboost" in python

Running behind a nginx proxy

Sometimes it is useful to run the Jupyter instance behind a nginx proxy, for instance:

  • you would prefer to access the notebook at a server URL with a path ( rather than a port (

  • you may have many different services in addition to Jupyter running on the same server, and want to nginx to help improve server performance in manage the connections

Here is a quick example NGINX configuration to get started. You'll need a server, a .crt and .key file for your server, and docker & docker-compose installed. Then just download the files at that gist and run docker-compose up -d to test it out. Customize the nginx.conf file to set the desired paths and add other services.

Host volume mounts and notebook errors

If you are mounting a host directory as /home/jovyan/work in your container and you receive permission errors or connection errors when you create a notebook, be sure that the jovyan user (UID=1000 by default) has read/write access to the directory on the host. Alternatively, specify the UID of the jovyan user on container startup using the -e NB_UID option described in the Common Features, Docker Options section


Manpage installation

Most containers, including our Ubuntu base image, ship without manpages installed to save space. You can use the following dockerfile to inherit from one of our images to enable manpages:

# Choose your desired base image
ARG BASE_CONTAINER=jupyter/datascience-notebook:latest

USER root

# `/etc/dpkg/dpkg.cfg.d/excludes` contains several `path-exclude`s, including man pages
# Remove it, then install man, install docs
RUN rm /etc/dpkg/dpkg.cfg.d/excludes && \
    apt-get update --yes && \
    dpkg -l | grep ^ii | cut -d' ' -f3 | xargs apt-get install --yes --no-install-recommends --reinstall man && \
    apt-get clean && rm -rf /var/lib/apt/lists/*


Adding the documentation on top of an existing singleuser image wastes a lot of space and requires reinstalling every system package, which can take additional time and bandwidth; the datascience-notebook image has been shown to grow by almost 3GB when adding manapages in this way. Enabling manpages in the base Ubuntu layer prevents this container bloat. Just use previous Dockerfile with original ubuntu image as your base container:

# Ubuntu 20.04 (focal) from 2020-04-23
ARG BASE_CONTAINER=ubuntu:focal-20200423@sha256:238e696992ba9913d24cfc3727034985abd136e08ee3067982401acdc30cbf3f

For Ubuntu 18.04 (bionic) and earlier, you may also require to workaround for a mandb bug, which was fixed in mandb >=


RUN echo "MANPATH_MAP ${CONDA_DIR}/bin ${CONDA_DIR}/man" >> /etc/manpath.config && \
    echo "MANPATH_MAP ${CONDA_DIR}/bin ${CONDA_DIR}/share/man" >> /etc/manpath.config && \

Be sure to check the current base image in base-notebook before building.


We also have contributed recipes for using JupyterHub.

Use JupyterHub's dockerspawner

In most cases for use with DockerSpawner, given any image that already has a notebook stack set up, you would only need to add:

  1. install the jupyterhub-singleuser script (for the right Python)

  2. change the command to launch the single-user server

Swapping out the FROM line in the jupyterhub/singleuser Dockerfile should be enough for most cases.

Credit: Justin Tyberg, quanghoc, and Min RK based on docker-stacks/issues/124 and docker-stacks/pull/185

Containers with a specific version of JupyterHub

To use a specific version of JupyterHub, the version of jupyterhub in your image should match the version in the Hub itself.

FROM jupyter/base-notebook:33add21fab64
RUN pip install --quiet --no-cache-dir jupyterhub==1.4.1 && \
    fix-permissions "${CONDA_DIR}" && \
    fix-permissions "/home/${NB_USER}"

Credit: MinRK



A few suggestions have been made regarding using Docker Stacks with spark.

Using PySpark with AWS S3

Using Spark session for hadoop 2.7.3

import os
# !ls /usr/local/spark/jars/hadoop* # to figure out what version of hadoop
os.environ['PYSPARK_SUBMIT_ARGS'] = '--packages "org.apache.hadoop:hadoop-aws:2.7.3" pyspark-shell'

import pyspark
myAccessKey = input()
mySecretKey = input()

spark = pyspark.sql.SparkSession.builder \
        .master("local[*]") \
        .config("spark.hadoop.fs.s3a.access.key", myAccessKey) \
        .config("spark.hadoop.fs.s3a.secret.key", mySecretKey) \

df ="s3://myBucket/myKey")

Using Spark context for hadoop 2.6.0

import os
os.environ['PYSPARK_SUBMIT_ARGS'] = '--packages com.amazonaws:aws-java-sdk:1.10.34,org.apache.hadoop:hadoop-aws:2.6.0 pyspark-shell'

import pyspark
sc = pyspark.SparkContext("local[*]")

from pyspark.sql import SQLContext
sqlContext = SQLContext(sc)

hadoopConf = sc._jsc.hadoopConfiguration()
myAccessKey = input()
mySecretKey = input()
hadoopConf.set("fs.s3.impl", "org.apache.hadoop.fs.s3native.NativeS3FileSystem")
hadoopConf.set("fs.s3.awsAccessKeyId", myAccessKey)
hadoopConf.set("fs.s3.awsSecretAccessKey", mySecretKey)

df ="s3://myBucket/myKey")


Using Local Spark JARs

import os
os.environ['PYSPARK_SUBMIT_ARGS'] = '--jars /home/jovyan/spark-streaming-kafka-assembly_2.10-1.6.1.jar pyspark-shell'
import pyspark
from pyspark.streaming.kafka import KafkaUtils
from pyspark.streaming import StreamingContext
sc = pyspark.SparkContext()
ssc = StreamingContext(sc,1)
broker = "<my_broker_ip>"
directKafkaStream = KafkaUtils.createDirectStream(ssc, ["test1"], {"": broker})



If you'd like to use packages from, see for an example of how to specify the package identifier in the environment before creating a SparkContext.


Use jupyter/all-spark-notebooks with an existing Spark/YARN cluster

FROM jupyter/all-spark-notebook

# Set env vars for pydoop
ENV HADOOP_HOME /usr/local/hadoop-2.7.3
ENV JAVA_HOME /usr/lib/jvm/java-8-openjdk-amd64
ENV HADOOP_CONF_HOME /usr/local/hadoop-2.7.3/etc/hadoop
ENV HADOOP_CONF_DIR  /usr/local/hadoop-2.7.3/etc/hadoop

USER root
# Add proper open-jdk-8 not just the jre, needed for pydoop
RUN echo 'deb jessie-backports main' > /etc/apt/sources.list.d/jessie-backports.list && \
    apt-get update --yes && \
    apt-get install --yes --no-install-recommends -t jessie-backports openjdk-8-jdk && \
    rm /etc/apt/sources.list.d/jessie-backports.list && \
    apt-get clean && rm -rf /var/lib/apt/lists/* && \
# Add hadoop binaries
    wget && \
    tar -xvf hadoop-2.7.3.tar.gz -C /usr/local && \
    chown -R "${NB_USER}:users" /usr/local/hadoop-2.7.3 && \
    rm -f hadoop-2.7.3.tar.gz && \
# Install os dependencies required for pydoop, pyhive
    apt-get update --yes && \
    apt-get install --yes --no-install-recommends build-essential python-dev libsasl2-dev && \
    apt-get clean && rm -rf /var/lib/apt/lists/* && \
# Remove the example hadoop configs and replace
# with those for our cluster.
# Alternatively this could be mounted as a volume
    rm -f /usr/local/hadoop-2.7.3/etc/hadoop/*

# Download this from ambari / cloudera manager and copy here
COPY example-hadoop-conf/ /usr/local/hadoop-2.7.3/etc/hadoop/

# Spark-Submit doesn't work unless I set the following
RUN echo "spark.driver.extraJavaOptions -Dhdp.version=" >> /usr/local/spark/conf/spark-defaults.conf  && \
    echo " -Dhdp.version=" >> /usr/local/spark/conf/spark-defaults.conf && \
    echo "spark.master=yarn" >>  /usr/local/spark/conf/spark-defaults.conf && \
    echo "spark.hadoop.yarn.timeline-service.enabled=false" >> /usr/local/spark/conf/spark-defaults.conf && \
    chown -R "${NB_USER}:users" /usr/local/spark/conf/spark-defaults.conf && \
    # Create an alternative HADOOP_CONF_HOME so we can mount as a volume and repoint
    # using ENV var if needed
    mkdir -p /etc/hadoop/conf/ && \
    chown "${NB_USER}":users /etc/hadoop/conf/


# Install useful jupyter extensions and python libraries like :
# - Dashboards
# - PyDoop
# - PyHive
RUN pip install --quiet --no-cache-dir jupyter_dashboards faker && \
    jupyter dashboards quick-setup --sys-prefix && \
    pip2 install --quiet --no-cache-dir pyhive pydoop thrift sasl thrift_sasl faker && \
    fix-permissions "${CONDA_DIR}" && \
    fix-permissions "/home/${NB_USER}"

USER root
# Ensure we overwrite the kernel config so that toree connects to cluster
RUN jupyter toree install --sys-prefix --spark_opts="\
    --master yarn
    --deploy-mode client
    --driver-memory 512m
    --executor-memory 512m
    --executor-cores 1
    --conf spark.hadoop.yarn.timeline-service.enabled=false

Credit: britishbadger from docker-stacks/issues/369

Run Jupyter Notebook/Lab inside an already secured environment (i.e., with no token)

(Adapted from issue 728)

The default security is very good. There are use cases, encouraged by containers, where the jupyter container and the system it runs within, lie inside the security boundary. In these use cases it is convenient to launch the server without a password or token. In this case, you should use the script to launch the server with no token:

For jupyterlab:

docker run jupyter/base-notebook:33add21fab64 jupyter lab --LabApp.token=''

For jupyter classic:

docker run jupyter/base-notebook:33add21fab64 jupyter notebook --NotebookApp.token=''

Enable nbextension spellchecker for markdown (or any other nbextension)

NB: this works for classic notebooks only

# Update with your base image of choice
FROM jupyter/minimal-notebook:latest


RUN pip install --quiet --no-cache-dir jupyter_contrib_nbextensions && \
    jupyter contrib nbextension install --user && \
    # can modify or enable additional extensions here
    jupyter nbextension enable spellchecker/main --user && \
    fix-permissions "${CONDA_DIR}" && \
    fix-permissions "/home/${NB_USER}"


Enable Delta Lake in Spark notebooks

Please note that the Delta Lake packages are only available for Spark version > 3.0. By adding the properties to spark-defaults.conf, the user no longer needs to enable Delta support in each notebook.

FROM jupyter/pyspark-notebook:latest

RUN pip install --quiet --no-cache-dir delta-spark==${DELTA_CORE_VERSION} && \
     fix-permissions "${HOME}" && \
     fix-permissions "${CONDA_DIR}"

USER root

RUN echo 'spark.sql.extensions' >> "${SPARK_HOME}/conf/spark-defaults.conf" && \
    echo 'spark.sql.catalog.spark_catalog' >> "${SPARK_HOME}/conf/spark-defaults.conf"


# Trigger download of delta lake files
RUN echo "from pyspark.sql import SparkSession" > /tmp/ && \
    echo "from delta import *" >> /tmp/ && \
    echo "spark = configure_spark_with_delta_pip(SparkSession.builder).getOrCreate()" >> /tmp/ && \
    python /tmp/ && \
    rm /tmp/

Add Custom Font in Scipy notebook

The example below is a Dockerfile to load Source Han Sans with normal weight which is usually used for web.

FROM jupyter/scipy-notebook:latest

RUN PYV=$(ls "${CONDA_DIR}/lib" | grep ^python) && \
    MPL_DATA="${CONDA_DIR}/lib/${PYV}/site-packages/matplotlib/mpl-data" && \
    wget --quiet -P "${MPL_DATA}/fonts/ttf/" && \
    sed -i 's/' "${MPL_DATA}/matplotlibrc" && \
    sed -i 's/#font.sans-serif:/font.sans-serif: Source Han Sans CN,/g' "${MPL_DATA}/matplotlibrc" && \
    sed -i 's/#axes.unicode_minus: True/axes.unicode_minus: False/g' "${MPL_DATA}/matplotlibrc" && \
    rm -rf "/home/${NB_USER}/.cache/matplotlib" && \
    python -c 'import matplotlib.font_manager;print("font loaded: ",("Source Han Sans CN" in [ for f in matplotlib.font_manager.fontManager.ttflist]))'