feat(xray): add tunnel health monitor (#5480)

* feat(xray): add tunnel health monitor

* fix(tunnelmonitor): reuse netproxy client and init logger in tests

Replace the duplicated newHTTPClient/dialContextWithProxy with netproxy.NewHTTPClient, which centralises the http/https/socks5 handling and avoids the dial-goroutine connection leak on context cancellation. Cap failures at the threshold during cooldown so the counter stays a true consecutive-failure count. Add TestMain to initialise the logger and fix the nil-pointer panic in the success-after-failure path.

* fix(tunnelmonitor): observable recovery, signal headroom, and hardening

Address the remaining review findings on the tunnel health monitor:

- Recovery is now synchronous and observable: the callback calls
  server.RestartXray() directly and returns its error instead of just
  enqueuing SIGUSR1, so a failed restart no longer masks as success and
  arms the cooldown while the tunnel is still down.
- Give the OS signal channel headroom (buffer 8) so producers cannot
  starve a SIGTERM/SIGINT out of the single slot.
- Warn at startup when the monitor is enabled without a proxy, since the
  probe then measures host connectivity rather than the xray tunnel.
- Cap failures at the threshold in the nil-recover branch too, matching
  the cooldown cap.
- Document the XUI_TUNNEL_HEALTH_* vars in .env.example and the README.
- Add tests for status-code classification, Normalize bounds, New proxy
  scheme errors, the recovery-error and nil-recover paths, the cooldown
  cap, and Run context cancellation (coverage 90%).

---------

Co-authored-by: Sanaei <ho3ein.sanaei@gmail.com>

main.go

@@ -3,6 +3,7 @@
 package main
 
 import (
+	"context"
 	"flag"
 	"fmt"
 	"log"
@@ -18,6 +19,7 @@ import (
 	"github.com/mhsanaei/3x-ui/v3/internal/database"
 	"github.com/mhsanaei/3x-ui/v3/internal/logger"
 	"github.com/mhsanaei/3x-ui/v3/internal/sub"
+	"github.com/mhsanaei/3x-ui/v3/internal/tunnelmonitor"
 	"github.com/mhsanaei/3x-ui/v3/internal/util/crypto"
 	"github.com/mhsanaei/3x-ui/v3/internal/util/sys"
 	"github.com/mhsanaei/3x-ui/v3/internal/web"
@@ -91,7 +93,7 @@ func runWebServer() {
 		return
 	}
 
-	sigCh := make(chan os.Signal, 1)
+	sigCh := make(chan os.Signal, 8)
 	// Trap shutdown signals
 	signal.Notify(sigCh, syscall.SIGHUP, syscall.SIGTERM, sys.SIGUSR1, os.Interrupt)
 	global.SetRestartHook(func() {
@@ -100,6 +102,27 @@ func runWebServer() {
 		default:
 		}
 	})
+
+	var stopTunnelHealthMonitor context.CancelFunc
+	monitorCfg := tunnelmonitor.ConfigFromEnv()
+	if monitorCfg.Enabled {
+		if monitorCfg.ProxyURL == "" {
+			logger.Warning("Tunnel health monitor enabled without XUI_TUNNEL_HEALTH_PROXY: the probe measures host connectivity, not the xray tunnel, so failures will restart xray without fixing host network issues")
+		}
+
+		monitorCtx, cancel := context.WithCancel(context.Background())
+		stopTunnelHealthMonitor = cancel
+
+		monitor, err := tunnelmonitor.New(monitorCfg, func(_ context.Context) error {
+			logger.Warning("Tunnel health monitor threshold reached, restarting xray-core")
+			return server.RestartXray()
+		})
+		if err != nil {
+			logger.Warning("Tunnel health monitor disabled: ", err)
+		} else {
+			go monitor.Run(monitorCtx)
+		}
+	}
 	for {
 		sig := <-sigCh
 
@@ -142,6 +165,10 @@ func runWebServer() {
 			}
 
 		default:
+			if stopTunnelHealthMonitor != nil {
+				stopTunnelHealthMonitor()
+			}
+
 			// --- FIX FOR TELEGRAM BOT CONFLICT (409) on full shutdown ---
 			tgbot.StopBot()
 			// ------------------------------------------------------------