This is yet another experiment to run Space Engineers on Linux. It kind of works but performance is not very good, especially during autosaving.
- The container image built by Docker Compose is self-contained except for the three mounted directories (see instructions below).
- The Torch Server GUI is displayed locally by Xpra over SSH (no need for VNC, remote desktop or X forwarding).
- Ability to run container images (e.g. Podman or Docker).
- Enough storage space for the container image (about 5 GiB) and the data directory (about 6 GiB).
Build the container image with Docker Compose:
make buildExport and upload the image to the host that will run it. Then create the directories to be mounted:
mkdir --parents /path/to/data/{ssh_admin,ssh_etc,torch}In /path/to/data/ssh_admin, create an authorized_keys file with the SSH public keys
you want to accept for the authentication of administrators.
Create a container with the following parameters:
- Mounted directories:
/path/to/data/torchto/home/admin/torch/path/to/data/ssh_etcto/etc/ssh/path/to/data/ssh_adminto/home/admin/.ssh
- Published ports:
- TCP 22 for SSH (and the Torch Server GUI via Xpra).
- UDP 27016 for the game server.
Make sure those ports are not blocked by the host's firewall.
Connect to the container with SSH or Xpra:
SL_REMOTE_HOST=foo.example.net make connect-remote-xpraNote that this Makefile target is only an example. Feel free to modify the actual
connection command. For example, you may need to adjust the SSH command with --ssh.
The Torch Server can crash for many reasons, for example if you start the server without a world. In such a case, the server will restart automatically and you'll have to try again with a different configuration.
I got ideas from the following: