This is a proposed approach for how to take any existing UBI-based language image, and turn it into a Dev Spaces optimized container image. Dev Spaces/Eclipse Che have introduced us to this concept that a container platform can be used, not just to host production applications, but also as a place optimized for developers to write, test and commit code. By running a development workspace with an IDE inside of a container and serving it through the browser, we can extend all of the consistency, immutability, and portability benefits of our application runtime environments to inner-loop development.
However, there is one challenge we have to deal with when working in this mode -- our production container images have been optimized for production. That means that they only contain the minimal tooling & libraries required to run our application and nothing for the developer. In a developer workspace, we need to have a few more tools available to us to make our development experience as smooth and efficient as possible. Thus far we've solved this problem by managing completely separate images for the purpose of development and deployment. However this creates additional overhead in having to manage the build process for these two sets of images.
A better experience could be just taking our existing application runtime images, and layering on top a standard set of developer tools to produce developer-optimized versions of our production images.
This repo automatically publishes the following images that are available to use in Dev Spaces:
CEkit is a modular, YAML-based framework for generating Containerfile-driven images that can be built using Docker, Podman or Buildah. To acheive our developer-optimized image builds, we're going to use a 3 layered approach:
zsh and the Oh My ZSH! project.With CEkit, each layer that we add can be represented by a module that defines packages and setup steps for that layer.
To get started download and install CEkit.
To start off, let's just build our generic developer base image. This image will simply contain our developer base and custom tooling layers on top of a base
Developer Base Image:
cekit build podmanThe developer base image is defined in the image.yaml file which is automatically picked up by CEkit. The image.yaml defines the name, version (tag), base image and modules that will be used to compose the image.
CEKit also supports overrides. Overrides allow you to modify certain parts of a build to produce slightly different variations of an image without defining a completely separate build. For our purposes, we're going to use overrides to inject an alternate language image as our FROM image.
To build a Java OpenJDK 17 version of our image, we can run:
cekit build --overrides images/devspaces-openjdk-17.yaml podmanTo build a NodeJS 18 image, run:
cekit build --overrides images/devspaces-nodejs-18.yaml podmanIf you still have a need or desire to provide a raw Containerfile for people to build, CEkit has you covered there too. In addition to the build image, the cekit build command also outputs the full source for the image, including a generated containerfile, which can be built directly usin Docker, Podman or Buildah. For example, after running the NodeJS example above, I can also run:
podman build -t nodejs-containerfile target/imageThis is useful for supporting a mixed mode where you want to have a streamlined CI process for producing images for your project, but would also like to make it easy for consumers of your project to build the images themselves without extra tooling.
cekit build --overrides images/cekit-builder.yaml podman
cekit build --overrides images/devspaces-base.yaml podman
cekit build --overrides images/devspaces-openjdk-17.yaml podman
cekit build --overrides images/devspaces-nodejs-18.yaml podman
cekit build --overrides images/devspaces-nodejs-20.yaml podman
cekit build --overrides images/devspaces-java-node-combined.yaml podman
cekit build --overrides images/devspaces-nested-podman.yaml podman
cekit build --overrides images/devspaces-python-311.yaml podman