So, some concrete examples for my previous Boot2Docker rules post:
I’m modifying boot2docker to
- if present, auto-start an image named ‘dom0:latest’. This image then orchestrates the remainder of the system.
- my personal dom0 image starts sshd and the containers I want this system to auto-run.
- Set up a `home-volume` container, which I -volumes-from mount into all my development containers.
When I do some development, testing or production, it happens in containers, the base OS is pristine, and can be trivially updated (atm, i’m using boot from USB flash and SD Card).
Similarly, the dom0 container is also a bare busybox container, cloned from the filesystem of the boot2docker image itself.. I’m not ready for my end goal of doing this to my notebook and desktop – but then, this setup is only a few days old :).
This setup uses my detect existing /var/lib/docker on HD pull request , and the dom0-rootfs, dom0-base and dom0 images, and then from there, and initial dev image.
2 customisations I’ve made to the boot2docker are persisted on the HD – /var/lib/boot2docker/etc/hostname is set to something useful to me, and the optional /var/lib/boot2docker/bootlocal.sh script starts the dom0 container at boot.
When I need a set of containers started, I can create a tiny orchestration container that can talk to the docker daemon and thus start more containers, controlling how they interact with each other and the outside world.