resource limits and concurrency for events and req

app/static/examples/config.example.yml

@@ -10,6 +10,12 @@ server:
   max_connections: 100
   max_subscriptions_per_client: 10
 
+resource_limits:
+  cpu_cores: 2 # Limit the number of CPU cores the application can use
+  memory_mb: 1024 # Cap the maximum amount of RAM in MB the application can use
+  heap_size_mb: 512 # Set a limit on the Go garbage collector's heap size in MB
+  max_goroutines: 100 # Limit the maximum number of concurrently running Go routines
+
 rate_limit:
   ws_limit: 100 # WebSocket messages per second
   ws_burst: 200 # Allowed burst of WebSocket messages

config/types/resourceLimits.go

@@ -0,0 +1,8 @@
+package config
+
+type ResourceLimits struct {
+	CPUCores      int `yaml:"cpu_cores"`
+	MemoryMB      int `yaml:"memory_mb"`
+	HeapSizeMB    int `yaml:"heap_size_mb"`
+	MaxGoroutines int `yaml:"max_goroutines"`
+}

config/types/serverConfig.go

@@ -18,4 +18,5 @@ type ServerConfig struct {
 	KindWhitelist   KindWhitelistConfig   `yaml:"kind_whitelist"`
 	DomainWhitelist DomainWhitelistConfig `yaml:"domain_whitelist"`
 	Blacklist       BlacklistConfig       `yaml:"blacklist"`
+	ResourceLimits  ResourceLimits        `yaml:"resource_limits"`
 }

main.go

@@ -25,19 +25,22 @@ func main() {
 	utils.EnsureFileExists("relay_metadata.json", "app/static/examples/relay_metadata.example.json")
 
 	restartChan := make(chan struct{})
-	go utils.WatchConfigFile("config.yml", restartChan)
+	go utils.WatchConfigFile("config.yml", restartChan) // Critical goroutine
 
 	signalChan := make(chan os.Signal, 1)
 	signal.Notify(signalChan, syscall.SIGINT, syscall.SIGTERM)
 
 	var wg sync.WaitGroup
 	for {
-		wg.Add(1)
+		wg.Add(1) // Add to WaitGroup for the server goroutine
+
 		cfg, err := config.LoadConfig("config.yml")
 		if err != nil {
 			log.Fatal("Error loading config: ", err)
 		}
 
+		utils.ApplyResourceLimits(&cfg.ResourceLimits) // Apply limits once before starting the server
+
 		client, err := db.InitDB(cfg)
 		if err != nil {
 			log.Fatal("Error initializing database: ", err)
@@ -54,22 +57,22 @@ func main() {
 		}
 
 		mux := setupRoutes()
+
+		// Start the server
 		server := startServer(cfg, mux, &wg)
 
 		select {
 		case <-restartChan:
 			log.Println("Restarting server...")
-
-			// Close server before restart
-			server.Close()
-			wg.Wait()
-
+			server.Close() // Stop the current server instance
+			wg.Wait()      // Wait for the server goroutine to finish
 			time.Sleep(3 * time.Second)
+
 		case <-signalChan:
 			log.Println("Shutting down server...")
-			server.Close()
-			db.DisconnectDB(client)
-			wg.Wait()
+			server.Close()              // Stop the server
+			db.DisconnectDB(client)     // Disconnect from MongoDB
+			wg.Wait()                   // Wait for all goroutines to finish
 			return
 		}
 	}
@@ -97,13 +100,14 @@ func startServer(config *configTypes.ServerConfig, mux *http.ServeMux, wg *sync.
 	}
 
 	go func() {
-		defer wg.Done() // Notify that the server is done shutting down
+		defer wg.Done() // Notify that the server goroutine is done
 		fmt.Printf("Server is running on http://localhost%s\n", config.Server.Port)
 		err := server.ListenAndServe()
 		if err != nil && err != http.ErrServerClosed {
 			fmt.Println("Error starting server:", err)
 		}
 	}()
+
 	return server
 }
 

server/handlers/event.go

@@ -16,6 +16,7 @@ import (
 )
 
 func HandleEvent(ws *websocket.Conn, message []interface{}) {
+	utils.LimitedGoRoutine(func() {
 		if len(message) != 2 {
 			fmt.Println("Invalid EVENT message format")
 			response.SendNotice(ws, "", "Invalid EVENT message format")
@@ -47,6 +48,7 @@ func HandleEvent(ws *websocket.Conn, message []interface{}) {
 		HandleKind(context.TODO(), evt, ws, eventSize)
 
 		fmt.Println("Event processed:", evt.ID)
+	})
 }
 
 func HandleKind(ctx context.Context, evt relay.Event, ws *websocket.Conn, eventSize int) {
@@ -152,4 +154,3 @@ func determineCategory(kind int) string {
 		return "unknown"
 	}
 }
-

server/handlers/req.go

@@ -21,6 +21,7 @@ var subscriptions = make(map[string]relay.Subscription)
 var mu sync.Mutex // Protect concurrent access to subscriptions map
 
 func HandleReq(ws *websocket.Conn, message []interface{}, subscriptions map[string][]relay.Filter) {
+	utils.LimitedGoRoutine(func() {
 		if len(message) < 3 {
 			fmt.Println("Invalid REQ message format")
 			response.SendClosed(ws, "", "invalid: invalid REQ message format")
@@ -102,6 +103,7 @@ func HandleReq(ws *websocket.Conn, message []interface{}, subscriptions map[stri
 			response.SendClosed(ws, subID, "error: could not send EOSE")
 			return
 		}
+	})
 }
 
 // QueryEvents queries events from the MongoDB collection based on filters

server/utils/applyResourceLimits.go

@@ -0,0 +1,113 @@
+package utils
+
+import (
+	"log"
+	"runtime"
+	"runtime/debug"
+	"sync"
+	"time"
+
+	configTypes "grain/config/types"
+)
+
+var (
+	maxGoroutinesChan   chan struct{}
+	wg                  sync.WaitGroup
+	goroutineQueue      []func()
+	goroutineQueueMutex sync.Mutex
+)
+
+func ApplyResourceLimits(cfg *configTypes.ResourceLimits) {
+	// Set CPU cores
+	runtime.GOMAXPROCS(cfg.CPUCores)
+
+	// Set maximum heap size
+	if cfg.HeapSizeMB > 0 {
+		heapSize := int64(uint64(cfg.HeapSizeMB) * 1024 * 1024)
+		debug.SetMemoryLimit(heapSize)
+		log.Printf("Heap size limited to %d MB\n", cfg.HeapSizeMB)
+	}
+
+	// Start monitoring memory usage
+	if cfg.MemoryMB > 0 {
+		go monitorMemoryUsage(cfg.MemoryMB)
+		log.Printf("Max memory usage limited to %d MB\n", cfg.MemoryMB)
+	}
+
+	// Set maximum number of Go routines
+	if cfg.MaxGoroutines > 0 {
+		maxGoroutinesChan = make(chan struct{}, cfg.MaxGoroutines)
+		log.Printf("Max goroutines limited to %d\n", cfg.MaxGoroutines)
+	}
+}
+
+// LimitedGoRoutine starts a goroutine with limit enforcement
+func LimitedGoRoutine(f func()) {
+	// By default, all routines are considered critical
+	goroutineQueueMutex.Lock()
+	goroutineQueue = append(goroutineQueue, f)
+	goroutineQueueMutex.Unlock()
+	attemptToStartGoroutine()
+}
+
+func attemptToStartGoroutine() {
+	goroutineQueueMutex.Lock()
+	defer goroutineQueueMutex.Unlock()
+
+	if len(goroutineQueue) > 0 {
+		select {
+		case maxGoroutinesChan <- struct{}{}:
+			wg.Add(1)
+			go func(f func()) {
+				defer func() {
+					wg.Done()
+					<-maxGoroutinesChan
+					attemptToStartGoroutine()
+				}()
+				f()
+			}(goroutineQueue[0])
+
+			// Remove the started goroutine from the queue
+			goroutineQueue = goroutineQueue[1:]
+
+		default:
+			// If the channel is full, consider dropping the oldest non-critical goroutine
+			dropOldestNonCriticalGoroutine()
+		}
+	}
+}
+
+func dropOldestNonCriticalGoroutine() {
+	goroutineQueueMutex.Lock()
+	defer goroutineQueueMutex.Unlock()
+
+	if len(goroutineQueue) > 0 {
+		log.Println("Dropping the oldest non-critical goroutine to free resources.")
+		goroutineQueue = goroutineQueue[1:]
+		attemptToStartGoroutine()
+	}
+}
+
+func WaitForGoroutines() {
+	wg.Wait()
+}
+
+func monitorMemoryUsage(maxMemoryMB int) {
+	for {
+		var memStats runtime.MemStats
+		runtime.ReadMemStats(&memStats)
+
+		usedMemoryMB := int(memStats.Alloc / 1024 / 1024)
+		if usedMemoryMB > maxMemoryMB {
+			log.Printf("Memory usage exceeded limit: %d MB used, limit is %d MB\n", usedMemoryMB, maxMemoryMB)
+			debug.FreeOSMemory() // Attempt to free memory
+
+			// If memory usage is still high, attempt to drop non-critical goroutines
+			if usedMemoryMB > maxMemoryMB {
+				dropOldestNonCriticalGoroutine()
+			}
+		}
+
+		time.Sleep(1 * time.Second)
+	}
+}