[proxy, client] Bound embed client WireGuard per-Device memory (#5962)

proxy/cmd/proxy/cmd/debug.go

@@ -109,6 +109,22 @@ var debugStopCmd = &cobra.Command{
 	SilenceUsage: true,
 }
 
+var debugPerfCmd = &cobra.Command{
+	Use:          "perf <pool-cap>",
+	Short:        "Live-retune the tunnel buffer pool cap on all running clients",
+	Args:         cobra.ExactArgs(1),
+	RunE:         runDebugPerfSet,
+	SilenceUsage: true,
+}
+
+var debugRuntimeCmd = &cobra.Command{
+	Use:          "runtime",
+	Short:        "Show runtime stats (heap, goroutines, RSS)",
+	Args:         cobra.NoArgs,
+	RunE:         runDebugRuntime,
+	SilenceUsage: true,
+}
+
 var debugCaptureCmd = &cobra.Command{
 	Use:   "capture <account-id> [filter expression]",
 	Short: "Capture packets on a client's WireGuard interface",
@@ -159,6 +175,8 @@ func init() {
 	debugCmd.AddCommand(debugLogCmd)
 	debugCmd.AddCommand(debugStartCmd)
 	debugCmd.AddCommand(debugStopCmd)
+	debugCmd.AddCommand(debugPerfCmd)
+	debugCmd.AddCommand(debugRuntimeCmd)
 	debugCmd.AddCommand(debugCaptureCmd)
 
 	rootCmd.AddCommand(debugCmd)
@@ -220,6 +238,18 @@ func runDebugStop(cmd *cobra.Command, args []string) error {
 	return getDebugClient(cmd).StopClient(cmd.Context(), args[0])
 }
 
+func runDebugPerfSet(cmd *cobra.Command, args []string) error {
+	n, err := strconv.ParseUint(args[0], 10, 32)
+	if err != nil {
+		return fmt.Errorf("invalid value %q: %w", args[0], err)
+	}
+	return getDebugClient(cmd).PerfSet(cmd.Context(), uint32(n))
+}
+
+func runDebugRuntime(cmd *cobra.Command, _ []string) error {
+	return getDebugClient(cmd).Runtime(cmd.Context())
+}
+
 func runDebugCapture(cmd *cobra.Command, args []string) error {
 	duration, _ := cmd.Flags().GetDuration("duration")
 	forcePcap, _ := cmd.Flags().GetBool("pcap")

proxy/cmd/proxy/cmd/root.go

@@ -15,11 +15,22 @@ import (
 
 	"github.com/netbirdio/netbird/shared/management/domain"
 
+	"github.com/netbirdio/netbird/client/embed"
 	"github.com/netbirdio/netbird/proxy"
 	nbacme "github.com/netbirdio/netbird/proxy/internal/acme"
 	"github.com/netbirdio/netbird/util"
 )
 
+const (
+	// envPreallocatedBuffers caps the per-tunnel buffer pool. Zero (unset)
+	// keeps the upstream uncapped default.
+	envPreallocatedBuffers = "NB_PROXY_PREALLOCATED_BUFFERS"
+	// envMaxBatchSize overrides the per-tunnel batch size, which controls
+	// how many buffers each receive/TUN worker eagerly allocates. Zero
+	// (unset) keeps the platform default.
+	envMaxBatchSize = "NB_PROXY_MAX_BATCH_SIZE"
+)
+
 const DefaultManagementURL = "https://api.netbird.io:443"
 
 // envProxyToken is the environment variable name for the proxy access token.
@@ -148,6 +159,45 @@ func runServer(cmd *cobra.Command, args []string) error {
 
 	logger.Infof("configured log level: %s", level)
 
+	var wgPool, wgBatch uint64
+	var perf embed.Performance
+	if raw := os.Getenv(envPreallocatedBuffers); raw != "" {
+		n, err := strconv.ParseUint(raw, 10, 32)
+		if err != nil {
+			return fmt.Errorf("invalid %s %q: %w", envPreallocatedBuffers, raw, err)
+		}
+		wgPool = n
+		v := uint32(n)
+		perf.PreallocatedBuffersPerPool = &v
+		logger.Infof("tunnel preallocated buffers per pool: %d", n)
+	}
+	if raw := os.Getenv(envMaxBatchSize); raw != "" {
+		n, err := strconv.ParseUint(raw, 10, 32)
+		if err != nil {
+			return fmt.Errorf("invalid %s %q: %w", envMaxBatchSize, raw, err)
+		}
+		wgBatch = n
+		v := uint32(n)
+		perf.MaxBatchSize = &v
+		logger.Infof("tunnel max batch size override: %d", n)
+	}
+	if wgPool > 0 {
+		// Each bind recv goroutine (IPv4 + IPv6 + ICE relay) plus
+		// RoutineReadFromTUN eagerly reserves `batch` message buffers for
+		// the lifetime of the Device. A pool cap below that floor blocks
+		// the receive pipeline at startup.
+		batch := wgBatch
+		if batch == 0 {
+			batch = 128
+		}
+		const recvGoroutines = 4
+		floor := batch * recvGoroutines
+		if wgPool < floor {
+			logger.Warnf("%s=%d is below the eager-allocation floor (~%d for batch=%d); startup may deadlock",
+				envPreallocatedBuffers, wgPool, floor, batch)
+		}
+	}
+
 	switch forwardedProto {
 	case "auto", "http", "https":
 	default:
@@ -188,6 +238,7 @@ func runServer(cmd *cobra.Command, args []string) error {
 		CertLockMethod:           nbacme.CertLockMethod(certLockMethod),
 		WildcardCertDir:          wildcardCertDir,
 		WireguardPort:            wgPort,
+		Performance:              perf,
 		ProxyProtocol:            proxyProtocol,
 		PreSharedKey:             preSharedKey,
 		SupportsCustomPorts:      supportsCustomPorts,