From de2c9ce9b6a11e8325de385617cbc8dc2172928b Mon Sep 17 00:00:00 2001
From: youfak <youfak@126.com>
Date: Fri, 31 Jul 2026 13:07:37 +0800
Subject: [PATCH] feat: run gateway runtime reporter loop

---
 internal/gateway/controlplane/reporter.go     | 32 +++++++++++-
 .../gateway/controlplane/reporter_test.go     | 49 +++++++++++++++++++
 2 files changed, 80 insertions(+), 1 deletion(-)

diff --git a/internal/gateway/controlplane/reporter.go b/internal/gateway/controlplane/reporter.go
index 7690ba9..48d4237 100644
--- a/internal/gateway/controlplane/reporter.go
+++ b/internal/gateway/controlplane/reporter.go
@@ -134,10 +134,40 @@ func (reporter *RuntimeReporter) Report(ctx context.Context) error {
 	return nil
 }
 
+// Run keeps the Controller session alive through bounded, periodic Runtime
+// reports. The snapshot stream may start after Register, so an absent local
+// snapshot is a normal waiting state rather than a process failure.
+func (reporter *RuntimeReporter) Run(ctx context.Context) error {
+	registration, err := reporter.Register(ctx)
+	if err != nil {
+		return err
+	}
+	ticker := time.NewTicker(registration.HeartbeatInterval)
+	defer ticker.Stop()
+	for {
+		if err := reporter.reportIfSnapshot(ctx); err != nil {
+			return err
+		}
+		select {
+		case <-ctx.Done():
+			return ctx.Err()
+		case <-ticker.C:
+		}
+	}
+}
+
+func (reporter *RuntimeReporter) reportIfSnapshot(ctx context.Context) error {
+	err := reporter.Report(ctx)
+	if errors.Is(err, snapshot.ErrInvalidRuntimeReport) {
+		return nil
+	}
+	return err
+}
+
 func validateRegistration(workerID string, response *controlplanev1.RegisterWorkerResponse) (Registration, error) {
 	if response == nil || response.GetWorkerId() != workerID || !workerruntime.ValidIdentifier(response.GetSessionId()) ||
 		response.GetOwnershipEpoch() == 0 || response.GetHeartbeatInterval() == nil || response.GetMaxStaleAge() == nil ||
-		response.GetHeartbeatInterval().AsDuration() <= 0 || response.GetMaxStaleAge().AsDuration() <= 0 {
+		response.GetHeartbeatInterval().AsDuration() <= 0 || response.GetMaxStaleAge().AsDuration() < response.GetHeartbeatInterval().AsDuration() {
 		return Registration{}, ErrInvalidOptions
 	}
 	return Registration{
diff --git a/internal/gateway/controlplane/reporter_test.go b/internal/gateway/controlplane/reporter_test.go
index 1917e0a..01902e7 100644
--- a/internal/gateway/controlplane/reporter_test.go
+++ b/internal/gateway/controlplane/reporter_test.go
@@ -88,6 +88,39 @@ func TestRuntimeReporterRejectsInvalidState(t *testing.T) {
 	}
 }
 
+func TestRuntimeReporterRunWaitsForSnapshotThenReports(t *testing.T) {
+	store := snapshot.NewStore("cluster-a", "worker-a")
+	client := &runClient{runtime: make(chan *controlplanev1.ReportRuntimeRequest, 8)}
+	reporter, err := NewRuntimeReporter(client, store, Options{
+		WorkerID: "worker-a", InstanceID: "instance-a", Zone: "zone-a", ProtocolVersion: 1, Now: time.Now,
+	})
+	if err != nil {
+		t.Fatalf("NewRuntimeReporter(): %v", err)
+	}
+	ctx, cancel := context.WithCancel(context.Background())
+	result := make(chan error, 1)
+	go func() { result <- reporter.Run(ctx) }()
+	time.Sleep(15 * time.Millisecond)
+	select {
+	case request := <-client.runtime:
+		t.Fatalf("ReportRuntime() before snapshot = %+v", request)
+	default:
+	}
+	applySnapshot(t, store)
+	select {
+	case request := <-client.runtime:
+		if request.GetReportSequence() != 1 {
+			t.Fatalf("report sequence = %d, want 1", request.GetReportSequence())
+		}
+	case <-time.After(time.Second):
+		t.Fatal("ReportRuntime() was not called after applying a snapshot")
+	}
+	cancel()
+	if err := <-result; !errors.Is(err, context.Canceled) {
+		t.Fatalf("Run() error = %v, want context canceled", err)
+	}
+}
+
 type clientStub struct {
 	register        *controlplanev1.RegisterWorkerRequest
 	registration    *controlplanev1.RegisterWorkerResponse
@@ -118,3 +151,19 @@ func applySnapshot(t *testing.T, store *snapshot.Store) {
 		t.Fatalf("Apply(): %v", err)
 	}
 }
+
+type runClient struct {
+	runtime chan *controlplanev1.ReportRuntimeRequest
+}
+
+func (*runClient) RegisterWorker(context.Context, *controlplanev1.RegisterWorkerRequest) (*controlplanev1.RegisterWorkerResponse, error) {
+	return &controlplanev1.RegisterWorkerResponse{
+		WorkerId: "worker-a", SessionId: "session-a", OwnershipEpoch: 7,
+		HeartbeatInterval: durationpb.New(10 * time.Millisecond), MaxStaleAge: durationpb.New(time.Second),
+	}, nil
+}
+
+func (client *runClient) ReportRuntime(_ context.Context, request *controlplanev1.ReportRuntimeRequest) (*controlplanev1.ReportRuntimeResponse, error) {
+	client.runtime <- request
+	return &controlplanev1.ReportRuntimeResponse{AcceptedOwnershipEpoch: 7}, nil
+}