[FIXED] Micromamba and Dockerfile error: /bin/bash: activate: No such file or directory

Issue

Used to have a Dockerfile with conda and flask that worked locally but after installing pytorch had to switch to micromamba, the old CMD no longer works after switching to image mambaorg/micromamba:1-focal-cuda-11.7.1, this is how my Dockerfile looks like right now:

FROM mambaorg/micromamba:1-focal-cuda-11.7.1

WORKDIR /app

COPY environment.yml .

RUN micromamba env create --file environment.yml

EXPOSE 5001

COPY app.py .

CMD ["/bin/bash", "-c", "source activate env-app && flask run --host=0.0.0.0 --port=5001"]

Now I get error:

/bin/bash: activate: No such file or directory

but before when I was using conda it was working, I think the CMD command needs to be updated.

Solution

Assuming these 2 files:

environment.yml

name: testenv
channels:
  - conda-forge
dependencies:
  - python >= 3.9
  - flask

app.py

from flask import Flask

# Create a Flask application
app = Flask(__name__)


# Define a route for the root URL
@app.route("/")
def hello_world():
    return "Hello, World!"


if __name__ == "__main__":
    # Run the Flask app on the local development server
    app.run()

---

Then in your Dockerfile you don't activate the environmnent but use the path to environment, e.g. /opt/conda/envs/testenv/bin/flask (note the testenv in the path and in the environment.yaml):

Dockerfile

FROM mambaorg/micromamba:1-focal-cuda-11.7.1

WORKDIR /app

COPY environment.yml .

RUN micromamba env create --file environment.yml

EXPOSE 5001

COPY app.py .

CMD micromamba run -n testenv flask run --host=0.0.0.0 --port=5001

---

To build:

docker build . -f Dockerfile -t my/image:name

To run:

docker run --rm -i -t -p 5001:5001 my/image:name

Then navigating the browser to localhost:5001 you should see Hello World.

Answered By - Andrej Kesely