Minimize size of docker image - docker

Question

I am building a docker image for running yarn jobs.
In order to install yarn, I need curl to fetch the package repository. After installing yarn, I am not really interested in curl anymore so I purge it again.
But this has no effect on the resulting docker image size since the layer with curl installed is still and underlying image layer (as far as I understand docker images).
I am less interested in this specific case (curl and yarn) but in general how to minimize my docker image in such a scenario. How can I "purge" a no longer needed underlying layer in my docker image?
Example Dockerfile for reference:
FROM ubuntu:focal
# Updating and installing curl (not required in final image)
RUN apt update && apt install -y curl
# Using curl to install yarn
RUN curl -sS https://dl.yarnpkg.com/debian/pubkey.gpg | apt-key add - &&\
echo "deb https://dl.yarnpkg.com/debian/ stable main" | tee /etc/apt/sources.list.d/yarn.list && \
apt update && apt install -y yarn
# Doing cleanup (no positive effect on image size)
RUN apt purge -y curl && rm -rf /var/lib/apt/lists/* && apt autoremove -y && apt clean -y
EDIT:
Just for clarification:
ubuntu/focal on it's own is just 74 MB in image size.
After running apt update it's at 95 MB
After apt installing curl wget git it's at 198 MB
Even purging all these installations doesn't bring me back to the 74 MB
multi-stage builds are a nice concept which I will look into.
This question although is about wheather or not it is possible to reduce a single image size again.

Answer 1

You can build your image using a multi stage Dockerfile.
For example:
FROM ubuntu:focal AS building_stage
RUN apt update && apt install -y curl
RUN curl -sS https://dl.yarnpkg.com/debian/pubkey.gpg | apt-key add - &&\
echo "deb https://dl.yarnpkg.com/debian/ stable main" | tee /etc/apt/sources.list.d/yarn.list && \
apt update && apt install -y yarn
RUN yarn install # or whatever you want to do with yarn
FROM ubuntu:focal AS running_stage
COPY --from=building_stage /root/node_modules .
After building this Dockerfile, the final image doesn't contain either yarn and curl, but it has necessary files for your final image to run. I didn't know what you wanted to do with yarn, so I couldn't show a pure example from your sample, but multi stage builds are probably the thing you want to use.