To build Docker images, you first need a Dockerfile. A Dockerfile is a plain text file named Dockerfile and has no file extension.

FROM <baseimage>        # get them from https://hub.docker.com

# set the working directory inside the image

WORKDIR /app

# copy the source code from the host machine to the image
# (host: your development machine or your server)

# the first dot (.) is the current working directory (CWD)
# of the host, the second dot (.) is the CWD of the image
COPY . .

# run any shell command inside the image

RUN <command>          # e.g., [npm run build]

# does not really expose the port 80 (only documentation)
# so that we know which port to expose

EXPOSE 80              # we expose ports with the --publish flag

# this command gets executed when the container starts
CMD <command>          # e.g., ["node”, “server.js”]

# end

Multistage Dockerfiles are used to optimize Dockerfiles. One use case is to create a builder and a serve stage with separate base images. This strategy can make the final image smaller and have a lower attack because it has fewer system libraries. Each stage starts with FROM.

# with as, you can give the current stage a variable name
FROM <baseimage> as builder

# now do something, install dependencies, build your code
RUN <command>         # e.g., g++ main.cpp

# the second stage could use a smaller image
# small images are based on alpine or you can build
# FROM scratch (this is a Docker image)
# if you do not need any system libraries
FROM scratch as serve

# you can now copy files from the builder stage
# e.g., the binary file that you build in that stage

COPY --from=builder ./hello-world ./hello-world

# this command gets executed when the container starts
CMD ["./hello-world”]

# end