Start a rootless container runing systemd

[GongT]

Is this possible? (note again, it's about running systemd inside container, not run podman as a service)

[rhatdan]

Yes. Just define a quadlet for the user and activate it.

You might need to enable linger..

You could run a systemd based container with a quadlet defining a rootless container.

[sbrivio-rh]

I would suggest to start with --privileged and... just try doing it, debug errors one by one, or report them here. There will be some accesses to hardware devices going horribly wrong, but maybe they all have workarounds.

[GongT]

My image is simple:

FROM fedora:latest
RUN dnf install systemd systemd-networkd

Run it will print many service failed start, podman exec -l systemctl list-units show there are multiple start job, everything "seems" to work. Then after random time (~1min), systemd in container core dump.

I searched around and found someone use systemd in docker without any issue, and he's using archlinux base image.
I switched my base image from fedora to archlink, and it just works...

I thought it was systemd problem, but it isn't. After some test, I finally written this entrypoint.sh. I don't know why this works, LOL.

#!/bin/bash
shopt -s nullglob
echo "$container_uuid" >/etc/machine-id
echo "$container_uuid" >/run/machine-id

rm -rf /usr/lib/systemd/system/local-fs.target.wants /usr/lib/systemd/system/graphical.target.wants /usr/lib/systemd/system/multi-user.target.wants
rm -rf /etc/systemd/system/*.target.wants

cd /usr/lib/systemd/system/sysinit.target.wants
rm -f ./*.mount ./*.path ./systemd-firstboot.service ./systemd-machine-id-commit.service ./systemd-sysusers.service ./systemd-tpm2-* ./systemd-update-*


capsh --print
exec /usr/lib/systemd/systemd --system --log-target=console --show-status=yes --log-color=yes --crash-reboot=no