Granted, systemd doesn't have the whole "pull some random image built by $DEITY-knows-who" repository thing (yet) + some weird, ad-hoc DSL to specify building and modifying images. However, that's a feature for me.
Systemd instead encourages you to just use the normal system tools (dnf, yum, pacman, debootstrap, etc.) to install a distro to a directory or volume. Using the distro-provided mechanisms for validating what you installed is what was intended. It lets you use standard system tools for provisioning the storage, from the trivial (just don't bother with anything fancy), to more complicated (overlays, LVM thin provisioning, etc.). Just much more natural to me. You can use standard scripting tools for running commands to manipulating images too (inc easily running commands inside the container, even when not running / 'unbooted', via systemd-run) - lot saner.
On "(yet)": They seem to be working on some kind of image management support. Unfortunately, heavily tied to specific OS features. E.g., btrfs. Which is a shame. However, as long it doesn't affect the existing "as you want" of the existing pieces, that's fine.
Basically, for the benefit of anyone else like me: systemd-nspawn / machinectl ++, docker--.
systemd machinectl
Posted Jul 21, 2016 14:19 UTC (Thu) by mathstuf (subscriber, #69389) [Link]
Copyright © 2018, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds