After running this configuration for a while, we
noticed that the containers didn't get back up once and the services
failed with the following error:
Error: current system boot ID differs from cached boot ID; an unhandled reboot has occurred.
This is hard to reproduce and seems to be a timing issue. However,
the logs indicated another issue that this patch now solves:
* The ExecStartPost= indicated that the user session got stopped before
which is required or sdnotify=healthy. Add explicit ordering for
user@. This unfortunately requires a statically declared uid.
Closes#259770Closes#207050
The motivation for the former is to not execute the container as root,
so you don't have to `sudo -i` to perform podman management tasks.
The idea behind healthchecks is to be able to keep the unit in the
activating state until the container is healthy, only then then unit is
marked as active.
The following changes were necessary:
* Move the ctr-id into `/run/${containerName}` to make podman can
actually write to it since it's now in its RuntimeDirectory.
* Make `sdnotify` option configurable (`healthy` for healthchecks that
must pass, default remains `conmon`).
* Set Delegate=yes for `sdnotify=healthy` to make sure a rootless
container can actually talk to sd_notify[1].
* Add a warning that lingering must be enabled to have a `systemd --user`
instance running which is required for the cgroup support to work
properly.
* Added a testcase for rootless containers with both conmon and
healthchecks.
[1] https://github.com/containers/podman/discussions/20573#discussioncomment-7612481
This adds a new `imageStream` option that can be used in conjunction
with `pkgs.dockerTools.streamLayeredImage` so that the image archive
never needs to be materialized in the `/nix/store`. This greatly
improves the disk utilization for systems that use container images
built using Nix because they only need to store image layers instead of
the full image. Additionally, when deploying the new system and only
new layers need to be built/copied.
