fix(agent): poll Windows service state instead of fixed sleep during updates

sc stop/start are async on Windows - they return immediately without
waiting for the service to actually change state. Replace fixed 10s
sleep with active polling (500ms interval, 60s timeout) to guarantee
the service has stopped before replacing binaries and started before
health checking.

Author: yllada

agent/updater/updates/update.go

@@ -106,8 +106,6 @@ func runUpdateProcess(basePath string) error {
 		return fmt.Errorf("error stopping agent: %v", err)
 	}
 
-	time.Sleep(10 * time.Second)
-
 	// Migration: check if old naming convention exists and migrate to new naming
 	oldBinPath := filepath.Join(basePath, oldBin)
 	if !fs.Exists(oldBinPath) {
@@ -171,7 +169,6 @@ func rollbackAgent(currentBin, backupBin, basePath string) {
 	logger.Info("Rolling back agent to previous version...")
 
 	svc.Stop(config.SERV_AGENT_NAME)
-	time.Sleep(5 * time.Second)
 
 	os.Remove(filepath.Join(basePath, currentBin))
 	os.Rename(filepath.Join(basePath, backupBin), filepath.Join(basePath, currentBin))

agent/version.json

@@ -1,4 +1,4 @@
 {
-    "version": "11.1.5",
-    "updater_version": "1.0.4"
+    "version": "11.1.4",
+    "updater_version": "1.0.5"
 }

shared/svc/svc_windows.go

@@ -7,24 +7,59 @@ import (
 	"fmt"
 	"os/exec"
 	"strings"
+	"time"
 )
 
-// Start starts a Windows service by name.
+const (
+	pollInterval = 500 * time.Millisecond
+	stopTimeout  = 60 * time.Second
+	startTimeout = 60 * time.Second
+)
+
+// Start starts a Windows service by name and waits until it's running.
 func Start(serviceName string) error {
+	// Already running? Nothing to do
+	if running, _ := IsActive(serviceName); running {
+		return nil
+	}
+
 	cmd := exec.Command("sc", "start", serviceName)
-	if err := cmd.Run(); err != nil {
-		return fmt.Errorf("failed to start service %s: %w", serviceName, err)
+	cmd.Run() // Ignore error, we'll check actual state
+
+	// Poll until running or timeout
+	deadline := time.Now().Add(startTimeout)
+	for time.Now().Before(deadline) {
+		if running, _ := IsActive(serviceName); running {
+			return nil
+		}
+		time.Sleep(pollInterval)
 	}
-	return nil
+
+	return fmt.Errorf("timeout waiting for service %s to start", serviceName)
 }
 
-// Stop stops a Windows service by name.
+// Stop stops a Windows service by name and waits until it's stopped.
 func Stop(serviceName string) error {
+	// Already stopped? Nothing to do
+	status, _ := Status(serviceName)
+	if status == StatusStopped {
+		return nil
+	}
+
 	cmd := exec.Command("sc", "stop", serviceName)
-	if err := cmd.Run(); err != nil {
-		return fmt.Errorf("failed to stop service %s: %w", serviceName, err)
+	cmd.Run() // Ignore error, we'll check actual state
+
+	// Poll until stopped or timeout
+	deadline := time.Now().Add(stopTimeout)
+	for time.Now().Before(deadline) {
+		status, _ := Status(serviceName)
+		if status == StatusStopped {
+			return nil
+		}
+		time.Sleep(pollInterval)
 	}
-	return nil
+
+	return fmt.Errorf("timeout waiting for service %s to stop", serviceName)
 }
 
 // Restart restarts a Windows service by stopping and starting it.
@@ -37,12 +72,11 @@ func Restart(serviceName string) error {
 
 // IsActive checks if a Windows service is running.
 func IsActive(serviceName string) (bool, error) {
-	cmd := exec.Command("sc", "query", serviceName)
-	output, err := cmd.Output()
+	status, err := Status(serviceName)
 	if err != nil {
-		return false, fmt.Errorf("failed to query service %s: %w", serviceName, err)
+		return false, err
 	}
-	return strings.Contains(string(output), "RUNNING"), nil
+	return status == StatusRunning, nil
 }
 
 // Status returns the status of a Windows service.