With 0.9.0, we are shifting toward a final push to 1.0.
Prior versions were all about proving out the idea, and now that we’ve proven it works it’s time to clean it up and get it across the line. That means going through the entire codebase and ensuring it will be maintainable going forward, well tested, and working like we want.
You may have noticed it has been far too long since 0.8.1, and a good chunk of that time was me going down a rabbit hole of “can we embed k3s directly in your server?”.
I learned a lot about k3s internals, but that experiment ultimately failed. So, dispite my best efforts you will still see the
k3s binary in the lib dir. Maybe in 2.0….
Nevertheless, lots of good stuff has been done:
Project files are now YAML based
While I have a love/hate relationship with yaml, it is far more devops standard than the ini format we used in 0.8 and before. We no longer even have code for the old format, so you will have to convert.
Always a Docker-based Builder
Because the Ruckstack Builder CLI needs Docker for building services, we are now relying on Docker all the time.
We now have a docker image available at ghcr.io/ruckstack/ruckstack for using however you want, and the CLI executables are now that same Docker image with a lightweight launcher wrapped around them.
That means Docker is now required to run the Ruckstack Builder CLI.
NOTE: As before, the installed server still does not use Docker. Only the Builder does
Output messages and error handling have been cleaned up. There is even a fancy spinning animation for long-running steps.
There is also now a
--verbose flag for all commands that provides debug-level output of what is going on.
Compress docker layers
The installer now compresses the stored docker layers, reducing the installer size as well as installation footprint.
Now, the M1 chip is no longer the most exciting thing in an Apple fan’s life
The complete list of features added and issues resolved can be found at our github tracker