[improv]自动更新后，拉起新进程时可能失败，需要重试。在A4上执行StarAgent自动更新时，有一定几率无法自动重启。

NewLifeX · Nov 16, 2024 · 0eb5f38 · 0eb5f38
1 parent 41b57bb
commit 0eb5f38
Show file tree

Hide file tree

Showing 3 changed files with 35 additions and 10 deletions.
diff --git a/NewLife.Remoting/Clients/ClientBase.cs b/NewLife.Remoting/Clients/ClientBase.cs
@@ -841,7 +841,7 @@ private async Task CheckUpgrade(Object? data)
                     // 执行更新，解压缩覆盖文件
                     var rs = ug.Update();
 
-                    // 执行前置命令
+                    // 执行更新后命令
                     if (rs && !info.Executor.IsNullOrEmpty()) ug.Run(info.Executor);
                     _lastVersion = info.Version;
 
@@ -854,6 +854,7 @@ private async Task CheckUpgrade(Object? data)
         {
             span?.SetError(ex, null);
             //Log?.Error(ex.ToString());
+            this.WriteErrorEvent("Upgrade", $"更新失败！{ex.Message}");
             throw;
         }
 
@@ -870,12 +871,21 @@ protected virtual void Restart(Upgrade upgrade)
         var name = asm.GetName().Name;
         if (name.IsNullOrEmpty()) return;
 
-        // 重新拉起进程
+        // 重新拉起进程。对于大多数应用，都是拉起新进程，然后退出当前进程；对于星尘代理，通过新进程来重启服务。
         var args = Environment.GetCommandLineArgs();
         if (args == null || args.Length == 0) args = new String[1];
         args[0] = "-upgrade";
+        var gs = args.Join(" ");
 
-        var rs = upgrade.Run(name, args.Join(" "));
+        // 执行重启，如果失败，延迟后再次尝试
+        var rs = upgrade.Run(name, gs, 3_000);
+        if (!rs)
+        {
+            var delay = 3_000;
+            this.WriteInfoEvent("Upgrade", $"拉起新进程失败，延迟{delay}ms后重试");
+            Thread.Sleep(delay);
+            rs = upgrade.Run(name, gs, 1_000);
+        }
 
         if (rs)
         {

diff --git a/NewLife.Remoting/Clients/Upgrade.cs b/NewLife.Remoting/Clients/Upgrade.cs
@@ -221,7 +221,14 @@ void Restore(IDictionary<String, String> dic)
     }
 
     /// <summary>启动当前应用的新进程。当前进程退出</summary>
-    public Boolean Run(String name, String args)
+    public Boolean Run(String name, String args) => Run(name, args, 3_000);
+
+    /// <summary>启动当前应用的新进程。当前进程退出</summary>
+    /// <param name="name"></param>
+    /// <param name="args"></param>
+    /// <param name="msWait"></param>
+    /// <returns></returns>
+    public Boolean Run(String name, String args, Int32 msWait)
     {
         var file = "";
         if (Runtime.Windows || Runtime.Mono)
@@ -245,12 +252,20 @@ public Boolean Run(String name, String args)
         }
 
         WriteLog("拉起进程 {0} {1}", file, args);
-        var p = file.EndsWithIgnoreCase(".dll") ?
-            RunShell("dotnet", $"{file} {args}") :
-            RunShell(file, args);
+        try
+        {
+            var p = file.EndsWithIgnoreCase(".dll") ?
+                RunShell("dotnet", $"{file} {args}") :
+                RunShell(file, args);
 
-        // 如果进程在指定时间退出，说明启动失败
-        return p != null && (!p.WaitForExit(1000) || p.ExitCode == 0);
+            // 如果进程在指定时间退出，说明启动失败
+            return p != null && (!p.WaitForExit(msWait) || p.ExitCode == 0);
+        }
+        catch (Exception ex)
+        {
+            WriteLog("启动进程失败：{0}", ex.Message);
+            return false;
+        }
     }
 
     static Process? RunShell(String fileName, String args) => Process.Start(new ProcessStartInfo(fileName, args) { UseShellExecute = true });

diff --git a/Samples/IoTZero/Services/MyDeviceService.cs b/Samples/IoTZero/Services/MyDeviceService.cs
@@ -99,7 +99,7 @@ public MyDeviceService(IPasswordProvider passwordProvider, ICacheProvider cacheP
         //SetChildOnline(dv, ip);
 
         // 登录历史
-        WriteHistory(dv, source + "设备鉴权", true, $"[{dv.Name}/{dv.Code}]鉴权成功 " + inf.ToJson(false, false, false), ip);
+        WriteHistory(dv, source + "登录", true, $"[{dv.Name}/{dv.Code}]鉴权成功 " + inf.ToJson(false, false, false), ip);
 
         var rs = new LoginResponse
         {