Multi-stage builds

Intermediate Docker

Mike Metzger

Data Engineering Consultant

Single-stage builds

Typical Docker images are created using a single FROM command
Each addition to the source image adds space and makes its management
Consider an application that must be compiled prior to use
- You can add all the necessary components to the image, compile it, and then configure the final image for use
- This often leaves superfluous content in the image even if it is not used

FROM ubuntu
RUN apt update
RUN apt install gcc -y
...
RUN make
CMD ["data_app"]

Multi-stage builds

Multi-stage builds use multiple containers
Typically has one or more build stages
Final components are copied into a final container image
The build stages are then removed automatically
- Saving space and minimizing the size of the container image
Uses some additional syntax in the Dockerfile
- AS <alias>
- COPY --from=<alias>

Multi-stage build example

# Create initial build stage
FROM ubuntu AS stage1

# Install compiler and compile code
RUN apt install gcc -y
...
RUN make


# Start new stage to create final image
FROM alpine-base

# Copy from first stage to final
COPY --from=stage1 /data_app /data_app

# Run application on container start
CMD ["data_app"]

Let's practice!

Intermediate Docker