Developing Docker containers: my 2016 workflow
The build, run, inspect and publish loop I used for Dockerfiles in 2016, and the unresolved problem of comparing changes inside a running container.
Companion video on the issues I have with Docker
In an attempt to improve my previous Docker container update workflow, I want to write my notes on how I develop Dockerfiles et al.
First I find the source repository of the Docker image I’m interested in.
For example https://hub.docker.com/r/abiosoft/caddy/ → https://github.com/abiosoft/caddy-docker
I then fork that into my repo and clone it, into for example ~/tmp/caddy-docker.
Typically I would make changes to the Dockerfile. Then:
- Run a build
docker build -t WIP .
- Run it
docker run -d --name caddy -p 2015:2015 -t WIP - See it running with
docker ps - Run a shell like so:
docker exec -i -t caddy sh - Stop it like so
docker kill caddy - To start it again
docker start caddy - Remove it like so
docker rm caddy
Hopefully I then have a Dockerfile I like.
I then docker login and push my image to the Docker hub and perhaps test on
a production CoreOS machine.
When satisfied I commit the changes to my git repo and create a PR.
Unsolved issues of my workflow
I would like to git diff changes in the running Docker filesystem.
I would like to examine files in the Docker images without shelling in.