feat: enhance WebRTC port allocation with existing allocation checks and increase upload timeout

pkg/gateway/middleware.go

@@ -1054,8 +1054,14 @@ func (g *Gateway) handleNamespaceGatewayRequest(w http.ResponseWriter, r *http.R
 		proxyReq.Header.Set(HeaderInternalAuthNamespace, validatedNamespace)
 	}
 
+	// Use a longer timeout for upload paths (IPFS add can be slow for large files)
+	proxyTimeout := 30 * time.Second
+	if strings.HasPrefix(r.URL.Path, "/v1/storage/upload") || strings.HasPrefix(r.URL.Path, "/v1/storage/pin") {
+		proxyTimeout = 300 * time.Second
+	}
+
 	// Execute proxy request using shared transport for connection pooling
-	httpClient := &http.Client{Timeout: 30 * time.Second, Transport: g.proxyTransport}
+	httpClient := &http.Client{Timeout: proxyTimeout, Transport: g.proxyTransport}
 	resp, err := httpClient.Do(proxyReq)
 	if err != nil {
 		cb.RecordFailure()

pkg/namespace/cluster_manager_webrtc.go

@@ -331,6 +331,7 @@ func (cm *ClusterManager) getClusterNodesWithIPs(ctx context.Context, clusterID
 		FROM namespace_cluster_nodes ncn
 		JOIN dns_nodes dn ON ncn.node_id = dn.id
 		WHERE ncn.namespace_cluster_id = ?
+		GROUP BY ncn.node_id
 	`
 	if err := cm.db.Query(internalCtx, &rows, query, clusterID); err != nil {
 		return nil, err

pkg/namespace/webrtc_port_allocator.go

@@ -49,6 +49,13 @@ func (wpa *WebRTCPortAllocator) AllocateSFUPorts(ctx context.Context, nodeID, na
 	retryDelay := 100 * time.Millisecond
 
 	for attempt := 0; attempt < maxRetries; attempt++ {
+		// Re-check for existing allocation (handles read-after-write lag on retries)
+		if attempt > 0 {
+			if existing, err := wpa.GetSFUPorts(ctx, namespaceClusterID, nodeID); err == nil && existing != nil {
+				return existing, nil
+			}
+		}
+
 		block, err := wpa.tryAllocateSFUPorts(internalCtx, nodeID, namespaceClusterID)
 		if err == nil {
 			wpa.logger.Info("SFU ports allocated",
@@ -148,6 +155,13 @@ func (wpa *WebRTCPortAllocator) AllocateTURNPorts(ctx context.Context, nodeID, n
 	retryDelay := 100 * time.Millisecond
 
 	for attempt := 0; attempt < maxRetries; attempt++ {
+		// Re-check for existing allocation (handles read-after-write lag on retries)
+		if attempt > 0 {
+			if existing, err := wpa.GetTURNPorts(ctx, namespaceClusterID, nodeID); err == nil && existing != nil {
+				return existing, nil
+			}
+		}
+
 		block, err := wpa.tryAllocateTURNPorts(internalCtx, nodeID, namespaceClusterID)
 		if err == nil {
 			wpa.logger.Info("TURN ports allocated",

scripts/upgrade-nodes.sh

@@ -75,6 +75,10 @@ resolve_nodes "$@" | while IFS='|' read -r user host pass; do
     echo "[$i/$node_count] $user@$host"
     upgrade_node "$user" "$host" "$pass"
     echo "  ✓ Done"
+    if [ "$i" -lt "$node_count" ]; then
+        echo "  Waiting 30s before next node..."
+        sleep 30
+    fi
     echo ""
 done