Sparks
Spark is the agent that runs game containers on a Linux game host. It needs rootful Docker Engine and the Docker CLI. Campfire’s guided installer also needs systemd, curl, base64, and sudo access. Linux amd64 and arm64 packages are supported.
Install and enroll
- Create a Spark from Sparks in the panel. Use a game address players can reach.
- Open its enrollment helper and check the agent connection URL.
- Run the generated command on the game host. It installs Spark and its systemd
service and saves the enrolled identity in
/var/lib/spark. - Confirm that the Spark is connected, then open the game ports needed by players.
The command contains a single-use token valid for 15 minutes. Generate another if it expires. Generated panel certificates are handled by the helper; custom private certificates may require a CA file on the host.
Spark connects outbound to the panel’s agent TLS listener, normally port 8443.
It needs no inbound management port. Game traffic uses the host’s allocated game
ports, separately from the management connection.
Capacity and networking
Edit a Spark’s Resources & networking settings to define the capacity
available to game servers and the host port ranges they may reserve. CPU, memory,
and disk limits are allocation budgets; set a limit to 0 for no configured
limit. Leave enough capacity for the host operating system, Docker, backup
archives, and temporary restoration files rather than allocating the whole host.
Port allocations specify a bind address, TCP or UDP, and a port or range. Each server reserves host ports from these allocations while its Ember defines the container ports. Configure both protocols when the game requires them. Allocation does not open a firewall or configure router forwarding: make the selected host ports reachable separately.
The Spark’s game address is what players use to reach its servers; it is distinct from the panel’s agent connection URL and the interface Docker binds. Check it when moving servers between hosts. Review existing reservations before reducing capacity or removing port allocations.
Disk quotas and host storage
Disk usage is measured on Spark. Hard game-data limits require Linux 5.14 or newer
and XFS or ext4 with project quota accounting and enforcement enabled on the
filesystem holding SPARK_DATA/runtime. XFS needs the prjquota or pquota mount
option; ext4 also needs its project and quota filesystem features.
Prepare the filesystem before starting Spark, and restart Spark after changing quota support. Without supported quotas, usage is monitored but hard disk limits are unavailable.
Game-data allocations do not include backup archives, Docker images, or container writable layers. Host storage therefore needs separate capacity management.
Maintenance
| Action | When to use it |
|---|---|
| Update | Install the newer Spark build offered by the panel. |
| Force update | Reinstall the panel’s bundled Spark build even when versions match. |
| Re-enroll | Replace the Spark’s credentials and restart its service. |
Remote updates require a connected Spark supporting the selected action and use the current panel’s bundled package, not an independently selected release. Updating preserves credentials, configuration, game files, and backups; game containers keep running. Preserve the Spark data directory across service and host restarts.
Connection and feature troubleshooting
Inspect service logs with sudo journalctl -u spark -f. For connection problems,
check the configured agent address, outbound reachability, certificate trust,
and whether the Spark is enabled in the panel. An expired identity needs
re-enrollment. The Features view shows which operations a connected version supports.
Feature support is negotiated with the connected agent. A Partial support entry means only some operations in that feature group are supported; it is not confirmation that every related action is available. Update Spark when a required capability is missing, then check the reported features again. If remote update is unavailable, upgrade the binary manually.
Cleanup and removal
Use Prune to remove leftover Campfire installer containers, containers for deleted servers, unused images tracked by Campfire, and unused volumes labeled for Campfire. It preserves running and stopped game servers, their game files and backups, and other applications’ resources. It is not a general Docker system prune. Finish active server operations and transfers before confirming cleanup, and inspect the result for reclaimed resources.
Before restarting or updating Spark, let active installation, file, backup, and transfer operations finish. Move or delete its servers before removing a Spark from the panel, then stop or uninstall the service on the old host.