Game Server Fleet

Game Server Fleet (beta)

The fleet allocates dedicated game servers on demand instead of asking you to run them yourself. It is beta and not part of the GA feature set, so expect it to change.

Enabling it

The fleet is off by default. An operator turns it on with FEATURE_FLEET_ENABLED and picks a backend with FLEET_BACKEND:

• Docker runs game servers as local containers, which suits a single VPS.
• Agones runs them on Kubernetes with Agones for autoscaling.
• Plugin hands allocation to an out-of-tree gRPC plugin for custom infrastructure, wired up with FLEET_PLUGIN_DIR and GGSCALE_FLEET_PLUGIN.

See the Configuration Reference for the full backend settings.

Access control

Fleet allocation is gated twice. The API key needs the fleet scope, and the tenant needs the dedicated_servers entitlement. A caller missing either gets 403.

Allocating through matchmaking

The usual way to get a server is the fleet_allocation matchmaking mode. A matched group is handed a provisioned server's address in the ticket result. See Matchmaking.

Agones lifecycle

On the Agones backend, your server binary must follow the Agones SDK lifecycle: call Ready() on startup, answer Health() checks, and handle Shutdown(). Describe the deployment in a Fleet custom resource that sets the transport, health check delays, port policy, and the pinned image digest.

Full reference: the /v1/fleets/* operations. The k3s and Agones test stack lives in the bw-ops repo under dev/fleet-agones/.