Big size of python image in Docker

Yes. It is normal.

I see there is an answer mentioning alpine. Alpine linux is quite minimal and it is quite common to see people suggest alpine in production.

However, I recently read this article: https://pythonspeed.com/articles/base-image-python-docker-images/, and the article suggest to use slim instead of alpine.

To use python slim, you can simply put this in top of your Dockerfile

FROM python:3.8-slim-buster

In the following image, I use fastapi + sqlalchemy + pipenv, and the size is relatively small compared to your original Python image.

enter image description here


Alpine Linux is a very lean distro avaliable for Docker. Without Python, it's around 5MB. With Python I'm getting images between 60 and 120 MB. The following Dockerfile yields a 110 MB image.

FROM alpine:3.4

RUN apk --update add \
      build-base python-dev \
      ca-certificates python \
      ttf-droid \
      py-pip \
      py-jinja2 \
      py-twisted \
      py-dateutil \
      py-tz \
      py-requests \
      py-pillow \
      py-rrd && \
    pip install --upgrade arrow \
                          pymongo \
                          websocket-client \
                          XlsxWriter && \
    apk del build-base python-dev && \
    rm -rf /var/cache/apk/* && \
    adduser -D -u 1001 noroot

USER noroot

CMD ["/bin/sh"]

Also, it's very well mantained.

A word of warning, though. Alpine uses musl libc instead of glibc, and some Python modules rely on glibc, but this usually isn't a problem.

A bigger issue is, that because of this, manylinux wheels are not avaliable for Alpine, and therefore the modules need to be compiled upon installation (pip install). In some cases this can make a difference in build time between 20 seconds on Debian and 9 minutes or more on Alpine. The grpcio-module is notorious for that; it takes forever to compile.

There is a (somewhat unreliable) workaround where you tell Python that it is manylinux compatible.


I just checked on my machine the standard ubuntu:trusty image is 188 MB and the image with all python stuff is 480MB. I see 800MB images quite often, those are usually ones that contain some meaningful application.

However, these images are based on our private images the official Docker library image seems much larger for some reason. They are aware of this fact and are trying to reduce it. Look at the discussion on the subject here

If you need a bit smaller image try this one 'rouge8/pythons' it is about 100MB smaller.

rouge8/pythons latest … 680.3 MB

Keep in mind, docker images are arranged as a hierarchical layer structure. So if you reuse the same underlying base image for many containers the size that each individual container adds is quite small. It will only be the difference between the base plus whatever you added into specific container.


You can try the python:{version}-alpine version. It's much smaller:

>> docker image ls |grep python
python    3.6-alpine     89.4 MB
python    3.6            689 MB
python    3.5            689 MB
python    3.5.2          687 MB
python    3.4            833 MB
python    2.7            676 MB

At time of writing it looks like the official image supports -alpine on all python versions.

https://hub.docker.com/_/python/

Tags:

Python

Docker

Recent Posts