fix: non-blocking loader with retry backoff, 30s timeout

cmd/gateway/main.go

@@ -52,12 +52,9 @@ func main() {
 	if l, err := serviceadapter.NewLoader(registry, "api"); err != nil {
 		log.Printf("ServiceAdapter loader disabled: %v", err)
 	} else {
-		if err := l.Start(30 * time.Second); err != nil {
+		l.Start(30 * time.Second)
-			log.Printf("ServiceAdapter loader failed: %v", err)
+		loader = l
-		} else {
+		log.Printf("ServiceAdapter loader started (background)")
-			loader = l
-			log.Printf("ServiceAdapter loader started")
-		}
 	}
 
 	router := server.NewRouter(healthChecker, dispatcher, temporalHandler, upstreamHandler)

internal/serviceadapter/informer.go

@@ -54,7 +54,7 @@ func NewLoader(registry *Registry, namespace string) (*Loader, error) {
 		token:     string(tokenBytes),
 		baseURL:   "https://kubernetes.default.svc",
 		client: &http.Client{
-			Timeout: 10 * time.Second,
+			Timeout: 30 * time.Second,
 			Transport: &http.Transport{
 				TLSClientConfig: &tls.Config{RootCAs: pool},
 			},
@@ -63,19 +63,22 @@ func NewLoader(registry *Registry, namespace string) (*Loader, error) {
 	}, nil
 }
 
-// Start loads ServiceAdapters immediately, then polls every interval.
+// Start begins loading ServiceAdapters in the background. Non-blocking.
-func (l *Loader) Start(interval time.Duration) error {
+// Retries on failure so the gateway can start serving immediately.
-	if err := l.load(); err != nil {
+func (l *Loader) Start(interval time.Duration) {
-		// Retry once after 2s — handles transient "storage reinitializing" (429)
-		log.Printf("serviceadapter loader: first attempt failed (%v), retrying in 2s", err)
-		time.Sleep(2 * time.Second)
-		if err := l.load(); err != nil {
-			return fmt.Errorf("serviceadapter loader: %w", err)
-		}
-	}
-
-	// Background poll for changes
 	go func() {
+		// Initial load with retries (API server may be slow on startup)
+		for attempt := 1; ; attempt++ {
+			if err := l.load(); err != nil {
+				backoff := time.Duration(min(attempt*5, 30)) * time.Second
+				log.Printf("serviceadapter loader: attempt %d failed (%v), retry in %s", attempt, err, backoff)
+				time.Sleep(backoff)
+				continue
+			}
+			break
+		}
+
+		// Poll for changes
 		ticker := time.NewTicker(interval)
 		defer ticker.Stop()
 		for {
@@ -89,8 +92,6 @@ func (l *Loader) Start(interval time.Duration) error {
 			}
 		}
 	}()
-
-	return nil
 }
 
 // Stop stops the background poll.