flask-bootstrap | Ready-to-use Twitter-bootstrap for use in Flask

The summary is that eventlet (0.32.0) is currently not compatible with Python 3.10 because it tries to patch types that have become immutable in Python 3.10.

Like with your requirements, it is good practice to be more specific with your Docker dependencies too. Today using the tag 3 for the Python Docker image will give you 3.10.0, unless it is using a cache. In the future it could be a different version. Since there is a compatibility issue with Python 3.10, use Python 3.9 - the currently latest Python 3.9 Docker tag is 3.9.7.

i.e. it should work once you change your first line of the Dockerfile to:

The incompatibility was fixed in commit a2ebbd2 on Dec 21, 2017 so you need a later version of mysqlclient.

mysqlclient 1.3.13 was released on Jun 27, 2018. Try it or any later version.

There is already an Beanstalk application that's running EC2s with docker containers that work just fine with these requirements so I'm not sure why this is happening now.

As to "why this is happening now" – two things I can think of:

• The Pip version in your base image has been updated; newer versions are smarter about conflicting dependencies (in that they refuse to install package constellations that might/should not work).
• Because your requirements.txt isn't necessarily fully locked down; there are transitive dependencies that get installed that aren't listed in the file that have become incompatible with one another.

However, since you already do have working container images, that's great! You could simply do

I've tried the following:

RUN pip install wheel

to your dockerfile before you attempt to install your requirements file.

Edit: I missed that virtual environments were being utilized here. I would argue that using a virtual environment is unnecessary in this case unless the OP is using their docker instance to run multiple python applications in parallel. There are cases to be made for using this pattern, though that does not appear to be the case here. As such, my suggestion would be to do away with venv altogether and simply install all dependencies inside the docker instance python installation, which would convert every venv/bin/pip call to a simple pip call.