[SOLVED] Docker provider: start a container with systemd enabled

Terraform Providers
1

I’m trying to start a Fedora 31 container which needs systemd to be enabled.
As an example, this works on the CLI to get things going:

docker run -d --privileged -v /sys/fs/cgroup:/sys/fs/cgroup:ro --name fc31 fedora:31 /usr/lib/systemd/systemd --system

File main.tf:

# Configure Docker provider and connect to the local Docker socket
provider "docker" {
  host = "unix:///var/run/docker.sock"
}

# Pull fedora:31
resource "docker_image" "fedora" {
  name = "fedora:31"
}

# Create a Fedora 31 container
resource "docker_container" "fedora" {
    depends_on = [
    docker_image.fedora,
  ]

  image = "fedora:31"
  name  = "fc31"

  privileged = true
  start = true
  must_run = true

  ports {
    internal = 80
    external = 80
  }

  upload {
    source = "${path.module}/deploy.sh"
    file   = "/deploy.sh"
  }

  command = ["/bin/bash", "-x", "/deploy.sh" ]

}

In the deploy.sh script being called are some systemctl-commands which fail:

2020-02-20T16:30:01.056848782Z + sudo systemctl daemon-reload
2020-02-20T16:30:01.069708041Z System has not been booted with systemd as init system (PID 1). Can't operate.
2020-02-20T16:30:01.069734469Z Failed to connect to bus: Host is down

Is it possible to start the container with systemd from the Terraform provider and if yes what should be added to main.tf?
TIA!

2

Answering my own question.

“Vanilla” Fedora 31 will not work. A F31 based image with systemd installed will.
Here I use the Docker image created in the official Cobbler CI-build.
Ref: https://github.com/cobbler/cobbler/blob/master/.travis.yml#L36

# Configure Docker provider and connect to the local Docker socket
provider "docker" {
  host = "unix:///var/run/docker.sock"
}

# Use cobbler:f31 (based on Fedora 31)
resource "docker_image" "cobbler" {
  name = "cobbler:f31"
}

# Custom Fedora 31 container to run Cobbler 3.x in
resource "docker_container" "cobbler" {
  image = "cobbler:f31"
  name  = "cobbler-f31"

  privileged = true
  must_run   = true

  volumes {
    container_path = "/sys/fs/cgroup"
    host_path      = "/sys/fs/cgroup"
    read_only      = true
  }

  ports {
    internal = 80
    external = 80
  }

  upload {
    source     = "${path.module}/deploy.sh"
    file       = "/deploy.sh"
    executable = true
  }

  entrypoint = ["/usr/lib/systemd/systemd", "--system"]
  #command    = ["/bin/bash", "-x", "/deploy.sh"]

}

ATM you’ll have to enter the container to run deploy.sh manually, the command won’t (yet).