Running systemd in a Concourse container

Question

I am currently trying to get a CentOS container working with systemd enabled in order to use as the environment for a Concourse task. The associated pipeline is building RPM packages that install services managed by systemd and I would like to run tests with the environment being as close as possible to the intended use case. My container is based on the short manual on using systemd within a CentOS container from the CentOS Docker Hub image description.

I am not 100% sure what the deciding issue is, but it seems to be related to one of the following circumstances (whether all or only some of these need a fix/workaround I am unsure of):

1. It seems to not be possible to bind-mount the host's (or rather the Concourse worker's because of it using Docker in Docker) cgroup filesystem into the CentOS container correctly, as Concourse has no configuration for bind-mounts for tasks. I have tried to generate the necessary filesystems "from scratch" inside the container, mounting a tmpfs on /sys/fs/cgroup and then mounting a cgroup filesystem named systemd on /sys/fs/cgroup/systemd, but I am unsure whether or not this is supposed to work.
2. Concourse uses the mounted-in garden-init binary as the image entrypoint and PID 1 which results in systemd running under a different PID. According to the CentOS manual, the entrypoint is supposed to be /usr/sbin/init, which is a symlink to /usr/lib/systemd/systemd While this should be possible with the --system flag, I still could not get everything to work completely.

I have tried different task configurations (run through a privileged fly execute with the commands being executed through bash -exc <command>) with the aim of being able to run systemctl is-system-running successfully, but haven't been able to do so. Configurations I have tried and the errors I have received with them are:

1. Running /usr/sbin/init with or without mounting the systemd cgroup from scratch and with or without running the command through exec results in the error Couldn't find an alternative telinit implementation to spawn..
2. Running /usr/lib/systemd/systemd --system without mounting the systemd cgroup from scratch before results in systemd failing to start up with the error message
   Cannot find cgroup mount point: No such file or directory
Failed to allocate manager object: No such file or directory
3. Running /usr/lib/systemd/systemd --system after mounting the systemd cgroup as described results in systemd apparently starting up correctly (the output is the same as when running everything outside of Concourse according to the CentOS documentation), but running systemctl is-system-running results in the error Failed to get D-Bus connection: Operation not permitted.

I did manage to get systemctl is-system-running to work in a manually created runc container by copying the container bundle created through Concourse/Garden and only editing the entrypoint (or rather the process.args field of the OCI config.json) to be /usr/sbin/init, so it seems like this is sufficient for my requirements. Unfortunately, I do not know how to configure my pipeline to achieve this for a container launched from Concourse's context.