fix(agones): make the kind gate's Agones lifecycle actually work

Several independent causes, all of which had to be right before the
Fleet could reach Ready.

The supervisor pointed --sdk-base-url at 127.0.0.1:9357, which is the
Agones sidecar's gRPC port; its HTTP surface is 9358, and that is what
AGONES_SDK_HTTP_PORT carries and what agones_sdk.gd reads. An HTTP
client against the gRPC port could never have worked, in kind or in
production.

The supervisor also treated the sidecar's first incomplete /gameserver
response as fatal. The sidecar accepts requests before the controller
populates status.address and status.ports, so this produced a restart
loop precisely during normal Agones startup. It now polls until the
endpoint is assigned or ReadyTimeout elapses.

server_boot.gd started ServerControl and the Agones SDK only under
--allocated-mode, but the kind smoke deliberately strips that flag, so
nothing served the readiness probe and the GameServer could never become
Ready. Lifecycle now keys on AGONES_SDK_HTTP_PORT, which Agones injects
into every managed container, while allocation and roster semantics stay
tied to --allocated-mode. The SDK node is added to the tree
non-deferred, since start_health() creates a Timer immediately.

Fleet: Agones assigns its own SDK service account and masks that token
from the game container while keeping it for the injected sidecar, so
the manifest must not pin serviceAccountName or
automountServiceAccountToken. Godot stores user:// under HOME, so HOME
points at the writable runtime volume to keep the root filesystem
read-only, and fsGroup makes that volume writable for the non-root user.

Namespace: Agones' Dynamic port policy injects a hostPort, which both
the baseline and restricted Pod Security Standards forbid, so the
workload namespace enforces privileged while continuing to audit and
warn against restricted.

NetworkPolicy: the injected sidecar reaches the Kubernetes API over
HTTPS, and NetworkPolicy applies to the whole Pod rather than to the
container whose token was masked.

The kind runner creates the namespace before Helm so Agones can install
its per-namespace SDK RBAC, scopes gameservers.namespaces to it, forces
the allocator and ping Services to ClusterIP because LoadBalancer
ingress never becomes ready in plain kind, and labels the node so the
production Fleet's on-demand/zone constraints are exercised rather than
edited out of the rendered manifest.

Game/scripts/server_boot.gd

@@ -73,6 +73,29 @@ func _ready() -> void:
 		printerr("cosmic-clash-server: allocated transport '%s' is not supported by this build" % assigned_transport)
 		get_tree().quit(1)
 		return
+	# Agones injects its HTTP port into every managed game-server container.
+	# Keep lifecycle readiness and health active in the reduced kind smoke even
+	# though that environment intentionally omits allocation/roster semantics.
+	var agones_managed := not OS.get_environment("AGONES_SDK_HTTP_PORT").is_empty()
+	if allocated_mode or agones_managed:
+		_control = ServerControlScript.new()
+		_control.name = "ServerControl"
+		_control.drain_requested.connect(_on_drain_requested)
+		_control.initial_connect_ready.connect(_on_initial_connect_ready)
+		get_tree().root.add_child.call_deferred(_control)
+		var control_err := _control.start(int(config.get_value("readiness-port")), OS.get_environment(String(config.get_value("drain-token-env"))))
+		if control_err != OK:
+			printerr("cosmic-clash-server: refusing to start with invalid readiness control port")
+			get_tree().quit(1)
+			return
+	if agones_managed:
+		_agones = AgonesSDKScript.new()
+		_agones.name = "AgonesSDK"
+		# Health creates and starts a Timer immediately, so the SDK node must be
+		# in the tree before start_health() runs.
+		get_tree().root.add_child(_agones)
+		if _agones.configure_from_environment():
+			_agones.start_health()
 	if allocated_mode:
 		var roster_file := String(config.get_value("join-authorisations-file"))
 		var key_file := String(config.get_value("join-authorisations-key-file"))
@@ -88,25 +111,10 @@ func _ready() -> void:
 			printerr("cosmic-clash-server: refusing to start with invalid join-authorisations-file")
 			get_tree().quit(1)
 			return
-		_control = ServerControlScript.new()
 		# An allocated process owns exactly the roster issued for this match.
 		# Never let the general-purpose direct-server default (one player) start
 		# an allocated match with only a partial assignment admitted.
 		config.values["min-players"] = required_min_players(true, roster_tokens.size(), int(config.get_value("min-players")))
-		_control.name = "ServerControl"
-		_control.drain_requested.connect(_on_drain_requested)
-		_control.initial_connect_ready.connect(_on_initial_connect_ready)
-		get_tree().root.add_child.call_deferred(_control)
-		var control_err := _control.start(int(config.get_value("readiness-port")), OS.get_environment(String(config.get_value("drain-token-env"))))
-		if control_err != OK:
-			printerr("cosmic-clash-server: refusing to start with invalid readiness control port")
-			get_tree().quit(1)
-			return
-		_agones = AgonesSDKScript.new()
-		_agones.name = "AgonesSDK"
-		get_tree().root.add_child.call_deferred(_agones)
-		if _agones.configure_from_environment():
-			_agones.start_health()
 		_connection_leases = ConnectionLeaseClientScript.new()
 		_connection_leases.name = "ConnectionLeases"
 		var lease_url := OS.get_environment("COSMIC_CLASH_CONTROL_PLANE_URL")

Game/scripts/server_control.gd

@@ -1,9 +1,10 @@
 class_name ServerControl
 extends Node
 
-# Small loopback HTTP control surface for allocated servers. The Go supervisor
-# uses GET /ready as the explicit process-ready probe and POST /drain during a
-# controlled termination. Direct/community servers do not start this node.
+# Small loopback HTTP control surface for lifecycle-managed servers. The Go
+# supervisor uses GET /ready as the explicit process-ready probe and POST
+# /drain during a controlled termination. Direct/community servers outside
+# Agones do not start this node.
 
 signal drain_requested
 signal initial_connect_ready

deploy/k8s/base/fleet.yaml

@@ -43,19 +43,22 @@ spec:
               labelSelector:
                 matchLabels:
                   app.kubernetes.io/name: game-server
-          serviceAccountName: match-server
-          automountServiceAccountToken: false
+          # Leave serviceAccountName unset: Agones assigns its SDK account and
+          # masks that account's token from this public game-server container,
+          # while retaining it in the injected SDK sidecar that needs API access.
           securityContext:
             runAsNonRoot: true
             runAsUser: 10001
             runAsGroup: 10001
+            fsGroup: 10001
+            fsGroupChangePolicy: OnRootMismatch
             seccompProfile:
               type: RuntimeDefault
           containers:
             - name: game-server
               image: ghcr.io/cosmic-clash/game-server@sha256:0000000000000000000000000000000000000000000000000000000000000000
               args:
-                - --sdk-base-url=http://127.0.0.1:9357
+                - --sdk-base-url=http://127.0.0.1:9358
                 - --ready-url=http://127.0.0.1:7780/ready
                 - --drain-url=http://127.0.0.1:7780/drain
                 - --initial-connect-ready-url=http://127.0.0.1:7780/initial-connect-ready
@@ -86,6 +89,10 @@ spec:
                 - --join-authorisations-key-file=/run/secrets/cosmic-clash/join-signing-keys.json
                 - --readiness-port=7780
               env:
+                # Godot stores user:// beneath HOME. Point it at the writable
+                # runtime volume while retaining a read-only root filesystem.
+                - name: HOME
+                  value: /run/cosmic-clash
                 - name: COSMIC_CLASH_SERVER_ID
                   valueFrom:
                     fieldRef:

deploy/k8s/base/namespace.yaml

@@ -3,7 +3,10 @@ kind: Namespace
 metadata:
   name: cosmic-clash
   labels:
-    pod-security.kubernetes.io/enforce: restricted
+    # Agones' Dynamic port policy injects a hostPort into every GameServer
+    # Pod. Kubernetes' built-in baseline and restricted policies both forbid
+    # host ports, so this workload namespace must enforce privileged while
+    # continuing to surface restricted-policy deviations in audit and warnings.
+    pod-security.kubernetes.io/enforce: privileged
     pod-security.kubernetes.io/audit: restricted
     pod-security.kubernetes.io/warn: restricted
-

deploy/k8s/base/network-policies.yaml

@@ -90,6 +90,12 @@ spec:
       ports:
         - protocol: TCP
           port: 8080
+    # The injected Agones SDK sidecar updates its GameServer through the
+    # kubernetes.default HTTPS Service. Its token is masked from the public
+    # game-server container by Agones, but NetworkPolicy applies to the Pod.
+    - ports:
+        - protocol: TCP
+          port: 443
     - ports:
         - protocol: UDP
           port: 53

deploy/k8s/base/service-accounts.yaml

@@ -7,13 +7,6 @@ automountServiceAccountToken: false
 ---
 apiVersion: v1
 kind: ServiceAccount
-metadata:
-  name: match-server
-  namespace: cosmic-clash
-automountServiceAccountToken: false
----
-apiVersion: v1
-kind: ServiceAccount
 metadata:
   name: allocator
   namespace: cosmic-clash

scripts/verify_kind_agones.sh

@@ -34,6 +34,8 @@ dump_cluster_state() {
   for ns in agones-system cosmic-clash; do
     echo "=== namespace ${ns}: pods ===" >&2
     kubectl -n "$ns" get pods -o wide >&2 2>&1 || true
+    echo "=== namespace ${ns}: services ===" >&2
+    kubectl -n "$ns" get services -o wide >&2 2>&1 || true
     # Events explain scheduling/image/probe failures that pod status alone
     # does not: FailedScheduling, ImagePullBackOff, readiness probe errors.
     echo "=== namespace ${ns}: recent events ===" >&2
@@ -98,15 +100,22 @@ fi
 kind create cluster --name "$cluster_name" --image "$kind_node_image" --wait 120s
 kind load docker-image "$game_server_image" --name "$cluster_name"
 
+# Agones creates its SDK service account and namespaced RBAC in each configured
+# GameServer namespace. The namespace must therefore exist before Helm runs.
+kubectl apply -f deploy/k8s/base/namespace.yaml
+
 helm repo add agones https://agones.dev/chart/stable >/dev/null
 helm repo update >/dev/null
 # Agones 1.49 otherwise requests 10,100 MiB of ephemeral storage for both its
 # controller and extensions pods, which exceeds a default single-node kind
-# cluster before the Fleet can be exercised. These are smoke-only bounds;
-# production resource sizing remains deployment-owned.
+# cluster before the Fleet can be exercised. Its allocator and ping Services
+# also default to LoadBalancer, whose ingress never becomes ready in plain kind.
+# These are smoke-only bounds; production sizing and exposure remain
+# deployment-owned.
 helm upgrade --install agones agones/agones \
   --namespace agones-system --create-namespace \
   --version "$agones_version" \
+  --set 'gameservers.namespaces[0]=cosmic-clash' \
   --set agones.crds.cleanup.enabled=true \
   --set agones.controller.replicas=1 \
   --set agones.controller.resources.requests.ephemeral-storage=128Mi \
@@ -115,6 +124,9 @@ helm upgrade --install agones agones/agones \
   --set agones.extensions.resources.requests.ephemeral-storage=128Mi \
   --set agones.extensions.resources.limits.ephemeral-storage=512Mi \
   --set agones.allocator.replicas=1 \
+  --set agones.allocator.service.serviceType=ClusterIP \
+  --set agones.ping.http.serviceType=ClusterIP \
+  --set agones.ping.udp.serviceType=ClusterIP \
   --wait --timeout 5m
 
 kubectl wait --for=condition=available deployment/agones-controller \
@@ -122,6 +134,14 @@ kubectl wait --for=condition=available deployment/agones-controller \
 kubectl wait --for=condition=available deployment/agones-allocator \
   -n agones-system --timeout=180s
 
+# The production Fleet only schedules on explicitly on-demand, zoned nodes.
+# Give the disposable node equivalent labels so this gate exercises those
+# constraints instead of rewriting them out of the rendered Fleet.
+kubectl label nodes --all \
+  cosmic-clash.io/capacity-type=on-demand \
+  topology.kubernetes.io/zone=kind-smoke \
+  --overwrite
+
 # The base Fleet intentionally carries a release-time digest placeholder. For
 # this isolated run only, replace that exact placeholder with the image loaded
 # into kind. No repository manifest is modified and no mutable image is used
@@ -139,7 +159,6 @@ sed -e "s|ghcr.io/cosmic-clash/game-server@sha256:${zero_digest}|$game_server_im
     -e '/- --allocated-mode$/d' \
   deploy/k8s/base/fleet.yaml > "$work_dir/fleet.yaml"
 
-kubectl apply -f deploy/k8s/base/namespace.yaml
 kubectl -n cosmic-clash create secret generic cosmic-clash-game-server \
   --from-literal=drain-token=kind-smoke-drain-token \
   --from-literal=join-signing-keys.json='{"kind-smoke-key":"a2luZC1zbW9rZS1zaWduaW5nLWtleQ=="}' \

server/security/test_fleet_manifests.py

@@ -18,11 +18,19 @@ class FleetManifestTest(unittest.TestCase):
             "protocol: UDP", "containerPort: 7777", "replicas: 2",
         ):
             self.assertIn(label, fleet)
-        for hardening in ("runAsNonRoot: true", "automountServiceAccountToken: false", "readOnlyRootFilesystem: true", "allowPrivilegeEscalation: false"):
+        for hardening in (
+            "runAsNonRoot: true", "readOnlyRootFilesystem: true",
+            "allowPrivilegeEscalation: false", "fsGroup: 10001",
+            "name: HOME", "value: /run/cosmic-clash",
+        ):
             self.assertIn(hardening, fleet)
+        # Agones must assign its SDK service account so it can keep the token
+        # available to its injected sidecar while masking it from the game.
+        self.assertNotIn("serviceAccountName:", fleet)
+        self.assertNotIn("automountServiceAccountToken:", fleet)
         for runtime in (
             "ghcr.io/cosmic-clash/game-server@sha256:",
-            "--sdk-base-url=http://127.0.0.1:9357",
+            "--sdk-base-url=http://127.0.0.1:9358",
             "--control-plane-url=http://control-plane.cosmic-clash.svc.cluster.local:8080",
             "--protocol-version=1",
             "/opt/cosmic-clash/CosmicClashServer.x86_64",
@@ -88,13 +96,23 @@ class FleetManifestTest(unittest.TestCase):
         base = self.read("base/kustomization.yaml")
         for field in ("kind: Service", "name: control-plane", "port: 8080", "targetPort: http"):
             self.assertIn(field, service)
-        for field in ("name: game-server-allowed-egress", "app.kubernetes.io/name: game-server", "port: 8080"):
-            self.assertIn(field, network)
+        game_server_egress = network.split("name: game-server-allowed-egress", 1)[-1].split("---", 1)[0]
+        for field in ("app.kubernetes.io/name: game-server", "port: 8080", "port: 443"):
+            self.assertIn(field, game_server_egress)
         self.assertIn("control-plane-service.yaml", base)
 
     def test_kind_runner_is_explicitly_separate_from_production_roster_flow(self):
         runner = (ROOT / "scripts/verify_kind_agones.sh").read_text()
         self.assertIn("Agones lifecycle smoke", runner)
+        self.assertIn("gameservers.namespaces[0]=cosmic-clash", runner)
+        for service in (
+            "agones.allocator.service.serviceType=ClusterIP",
+            "agones.ping.http.serviceType=ClusterIP",
+            "agones.ping.udp.serviceType=ClusterIP",
+        ):
+            self.assertIn(service, runner)
+        self.assertIn("cosmic-clash.io/capacity-type=on-demand", runner)
+        self.assertIn("topology.kubernetes.io/zone=kind-smoke", runner)
         self.assertIn("--control-plane-url=", runner)
         self.assertIn("--allocated-mode", runner)
         validator = (ROOT / "scripts/verify_agones_allocation_response.py").read_text()

server/security/test_kubernetes_policies.py

@@ -10,10 +10,12 @@ class KubernetesPolicyTest(unittest.TestCase):
     def read(self, name):
         return (BASE / name).read_text()
 
-    def test_namespace_enforces_restricted_pod_security(self):
+    def test_namespace_allows_agones_host_ports_and_audits_restricted(self):
         namespace = self.read("namespace.yaml")
-        for key in ("enforce", "audit", "warn"):
+        self.assertIn("pod-security.kubernetes.io/enforce: privileged", namespace)
+        for key in ("audit", "warn"):
             self.assertIn(f"pod-security.kubernetes.io/{key}: restricted", namespace)
+        self.assertIn("Agones' Dynamic port policy", namespace)
 
     def test_workload_is_non_root_immutable_and_unprivileged(self):
         deployment = self.read("control-plane-deployment.yaml")

server/supervisor/supervisor.go

@@ -199,7 +199,7 @@ func (s *Supervisor) Start(ctx context.Context) error {
 	env := append([]string(nil), os.Environ()...)
 	env = append(env, s.config.Environment...)
 	if s.config.SDKBaseURL != "" {
-		port, address, err := s.assignedEndpoint(ctx)
+		port, address, err := s.waitAssignedEndpoint(ctx)
 		if err != nil {
 			return err
 		}
@@ -269,6 +269,30 @@ func (s *Supervisor) Start(ctx context.Context) error {
 	return nil
 }
 
+// waitAssignedEndpoint covers the short interval between the SDK sidecar
+// accepting requests and the GameServer controller populating status.address
+// and status.ports. Treating the sidecar's first incomplete response as fatal
+// creates a restart loop precisely while Agones is finishing normal startup.
+func (s *Supervisor) waitAssignedEndpoint(ctx context.Context) (int, string, error) {
+	deadline := time.NewTimer(s.config.ReadyTimeout)
+	defer deadline.Stop()
+	var lastErr error
+	for {
+		port, address, err := s.assignedEndpoint(ctx)
+		if err == nil {
+			return port, address, nil
+		}
+		lastErr = err
+		select {
+		case <-ctx.Done():
+			return 0, "", ctx.Err()
+		case <-deadline.C:
+			return 0, "", fmt.Errorf("assigned endpoint timed out: %w", lastErr)
+		case <-time.After(s.config.PollInterval):
+		}
+	}
+}
+
 func (s *Supervisor) signalInitialConnectReady(ctx context.Context) error {
 	if s.config.AdmissionURL == "" {
 		return nil

server/supervisor/supervisor_test.go

@@ -165,6 +165,44 @@ func TestAllocatedStartInjectsDynamicEndpointAndCallsReadyAfterProbe(t *testing.
 	}
 }
 
+func TestAllocatedStartWaitsForAgonesToAssignEndpoint(t *testing.T) {
+	requests := 0
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		switch r.URL.Path {
+		case "/gameserver":
+			requests++
+			if requests == 1 {
+				_, _ = w.Write([]byte(`{"status":{}}`))
+				return
+			}
+			_, _ = w.Write([]byte(`{"status":{"address":"203.0.113.9","ports":[{"name":"game","port":31001}]}}`))
+		case "/ready-probe", "/ready":
+			w.WriteHeader(http.StatusOK)
+		default:
+			w.WriteHeader(http.StatusNotFound)
+		}
+	}))
+	defer server.Close()
+
+	s, err := New(Config{
+		Command: []string{"/bin/sh", "-c", "true"}, SDKBaseURL: server.URL,
+		ReadyURL: server.URL + "/ready-probe", ReadyTimeout: time.Second,
+		PollInterval: time.Millisecond,
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err := s.Start(context.Background()); err != nil {
+		t.Fatal(err)
+	}
+	if err := s.Wait(); err != nil {
+		t.Fatal(err)
+	}
+	if requests < 2 {
+		t.Fatalf("gameserver requests = %d, want at least 2", requests)
+	}
+}
+
 func TestAllocatedStartMaterializesWorkloadAuthenticatedRosterBeforeChild(t *testing.T) {
 	rosterPath := filepath.Join(t.TempDir(), "join-roster.json")
 	sdk := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {