whaler

What? A command-line tool for visually investigating the disk usage of docker images

Why? Large images are slow to move and expensive to store. They cost developer productivity by lengthening devops tasks and often contain unnecessary data

Who is this for? Primarily for engineers working with images containing Python packages.

User Stories

This tool should allow you to answer questions such as: 1. Which file types are occupying the most disk space? 2. Which are my largest Python packages? 3. What are my unknown causes of high disk usage?

Quick start

pip install whaler

Run against a local directory

➜ whaler .venv
Running bash -c cd .venv && du -a -k
Done. Serving output at http://localhost:8000 (ctrl+c to exit)
Running python3 -m http.server 8000 --directory=_whaler/html

Run against a docker image

The tool will pull the image first if it is not present. ``` whaler --image='hl:latest' / Running docker run --rm --entrypoint=du --workdir=/ hl:latest -a -k Ignoring what seems to be non-fatal error(s): du: cannot access './proc/1/task/1/fd/4': No such file or directory du: cannot access './proc/1/task/1/fdinfo/4': No such file or directory du: cannot access './proc/1/fd/3': No such file or directory du: cannot access './proc/1/fdinfo/3': No such file or directory

Done. Serving output at http://localhost:8000 (ctrl+c to exit) Running python3 -m http.server 8000 --directory=_whaler/html ```

Play with a hosted demo

Limitations

1. Platform: whaler uses

   du

   to gather disk usage data. It must be present in your docker image
2. Scale: I have tested the web UI with up to 500,000 file system nodes with

   du

   output of up to ~100MB.

Alternatives/Complements to this tool:

1. Whaler can tell you what is taking up space in the final layer of your Docker image, but you may have intermediate layers which are contributing to the image size. For diving through the layers, use dive
   • Related: read up on multi-stage builds to understand how to mitigate the problem of intermediate layers bloating your image.
2. For investigating disk usage in non-docker directories, Disk Inventory X is a great tool on OS X which I have based whaler on.
3. GoogleContainerTools/distroless for base images not containing standard linux distro contents

See also

1. a good blog post on making small python images

Developing

See

.github/workflows/test.yml

For UI, see whaler-ui