Writing Dockerfile for a Python project the right way Serge Matveenko Twitter: @lig1 Github: @lig assaia.aero 

No content

Straightforward (good) FROM python:3.7 WORKDIR /usr/src/app COPY requirements.txt ./ RUN pip install --no-cache-dir -r requirements.txt COPY . . CMD [ "python", "./your-daemon-or-script.py" ] 

Straightforward (bad) FROM python:3 COPY . /usr/src/app/ RUN pip install --no-cache-dir -r requirements.txt WORKDIR /usr/src/app CMD python ./your-daemon-or-script.py 

Caching (obvious) FROM python:3.7 WORKDIR /usr/src/app COPY requirements.txt ./ RUN pip install --no-cache-dir -r requirements.txt COPY . . CMD [ "python", "./your-daemon-or-script.py" ] FROM python:3 COPY . /usr/src/app/ RUN pip install --no-cache-dir -r requirements.txt WORKDIR /usr/src/app CMD python ./your-daemon-or-script.py 

Caching (non-obvious) FROM python:3.7 WORKDIR /usr/src/app COPY requirements.txt ./ RUN pip install --no-cache-dir -r requirements.txt COPY . . CMD [ "python", "./your-daemon-or-script.py" ] FROM python:3 COPY . /usr/src/app/ RUN pip install --no-cache-dir -r requirements.txt WORKDIR /usr/src/app CMD python ./your-daemon-or-script.py 

Python version FROM python:3.7 WORKDIR /usr/src/app COPY requirements.txt ./ RUN pip install --no-cache-dir -r requirements.txt COPY . . CMD [ "python", "./your-daemon-or-script.py" ] FROM python:3 COPY . /usr/src/app/ RUN pip install --no-cache-dir -r requirements.txt WORKDIR /usr/src/app CMD python ./your-daemon-or-script.py 

Python version (even better) FROM python:3.7-stretch WORKDIR /usr/src/app COPY requirements.txt ./ RUN pip install --no-cache-dir -r requirements.txt COPY . . CMD [ "python", "./your-daemon-or-script.py" ] FROM python:3 COPY . /usr/src/app/ RUN pip install --no-cache-dir -r requirements.txt WORKDIR /usr/src/app CMD python ./your-daemon-or-script.py 

CMD FROM python:3.7 WORKDIR /usr/src/app COPY requirements.txt ./ RUN pip install --no-cache-dir -r requirements.txt COPY . . CMD [ "python", "./your-daemon-or-script.py" ] FROM python:3 COPY . /usr/src/app/ RUN pip install --no-cache-dir -r requirements.txt WORKDIR /usr/src/app CMD python ./your-daemon-or-script.py 

CMD (now you see me…) FROM python:3.7 WORKDIR /usr/src/app COPY requirements.txt ./ RUN pip install --no-cache-dir -r requirements.txt COPY . . ENTRYPOINT [ "python", "./your-daemon-or-script.py" ] CMD [ "run" ] FROM python:3 COPY . /usr/src/app/ RUN pip install --no-cache-dir -r requirements.txt WORKDIR /usr/src/app ENTRYPOINT python ./your-daemon-or-script.py CMD run 

That was too easy 

What if I want to build from …? 

FROM nvidia/cuda:9.0-cudnn7-runtime-ubuntu16.04 # :) FROM nvidia/cuda:9.0-cudnn7-runtime-ubuntu16.04 RUN apt-get update && apt-get install -y lib… lib… ENV PYTHON_VERSION=3.6.6 RUN curl -L https://github.com/pyenv/pyenv-installer/raw/master/bin/pyenv-installer | bash && \ pyenv install ${PYTHON_VERSION} && \ pyenv global ${PYTHON_VERSION} && \ curl -L https://raw.githubusercontent.com/kennethreitz/pipenv/master/get-pipenv.py | python && \ pipenv --python python COPY ./Pipfile ./Pipfile.lock ./ RUN pipenv install --deploy --dev 

Two different Tensorflows??? FROM nvidia/cuda:9.0-cudnn7-runtime-ubuntu16.04 AS base COPY ./Pipfile ./Pipfile.lock ./ RUN pipenv install --deploy --dev CMD ["pipenv", "run", "python", "-m", "my.worker"] FROM base AS base-gpu RUN pipenv run pip install tensorflow-gpu==$(pipenv run pip show tensorflow | grep '^Version' | awk '{print $2}') FROM base as app COPY . . FROM base-gpu as gpu COPY . . 

Questions? Serge Matveenko Twitter: @lig1 Github: @lig Telegram: @SergeMatveenko assaia.aero