graphit.py -i <yourlinuxkityml> -o <graphname>
Output is an SVG diagram showing:
- All Images (by name, image and tag) in the configuration.
- Colourcoding for init, onboot and service images.
- Mounts for onboot and service images.
- Source and dest information for each mount.
- Information on each mount (rbind etc).
An example is here: example.svg
Graphviz and associated tools will need to be installed on your system. for me:
brew install graphviz
You'll also need docker installed on the system and have the docker command in your path. we call out to it to inspect the LinuxKit images.
Needed Python3 dependancies are in requirements.txt
pip3 install -r requirements.txt
Working with LinuxKit now for a few weeks, debugging issues I found was a lot easier with a "map" of the system. This is a Very Hacky attempt at automating that.
This only visualizes binds from the LABEL metadata in each image's dockerfile (org.mobyproject.config). I know these are often overridden in the linuxkit yaml files, so we need to also read these in and display them.
I'm stll not 100% sure which ones "win" or how they are merged when both the LABEL and the yaml have binds.
The files section adds things to the "RootFS" just like images in init, i'll be adding these to the diagram next.
This is my first time playing with GraphViz and it shows. Making this tidier with more readable line labels and spacing would be awesome!
I'm currently calling out to docker inspect (and docker pull to have an image to inspect). This is nasty, however the pyDocker client doesnt seem to support inspect actions on anything other than running containers. I didn't have time to look into this.