fix(windows): prevent restart race from duplicate schtasks /Run

andyk-ms · BradGroux · commit 9a31bc85e7f8 · 2026-05-09T04:05:44.000-05:00
When CLI runs 'openclaw gateway restart', it does schtasks /End then schtasks /Run. Meanwhile the dying gateway fires a self-restart script (relaunchGatewayScheduledTask) that also retries schtasks /Run up to 12 times. Both succeed, creating duplicate gateway windows. Fix: the self-restart .cmd script now checks if the task is already running before attempting schtasks /Run. If another restart (CLI) beat it there, the script exits cleanly instead of spawning a duplicate. Fixes #52044
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -65,6 +65,7 @@ Docs: https://docs.openclaw.ai
 - Plugin skills/Windows: publish plugin-provided skill directories as junctions on Windows so standard users without Developer Mode can register plugin skills without symlink EPERM failures. Fixes #77958. (#77971) Thanks @hclsys and @jarro.
 - Shell env/Windows: hide the login-shell environment probe child window so gateway startup and shell-env refreshes do not flash a console on Windows. Fixes #78159. (#78266) Thanks @BradGroux.
 - MS Teams: surface blocked Bot Framework egress by logging JWKS fetch network failures and adding a Bot Connector send hint for transport-level reply failures. Fixes #77674. (#78081) Thanks @Beandon13.
+- Windows/restart: skip duplicate scheduled-task `/Run` calls when the gateway task is already running, using a locale-stable PowerShell task-state probe before retrying. Fixes #52044. (#52487) Thanks @andyk-ms.
 - Gateway/sessions: fast-path already-qualified model refs while building session-list rows so `openclaw sessions` and Control UI session lists avoid heavyweight model resolution on large stores. (#77902) Thanks @ragesaq.
 - Contributor PRs: remind external contributors to redact private information like IP addresses, API keys, phone numbers, and non-public endpoints from real behavior proof. Thanks @pashpashpash.
 - ACP bridge: relay Gateway exec approval prompts from active ACP turns to the ACP client's `session/request_permission` handler before resolving the Gateway approval. Thanks @amknight.
diff --git a/src/infra/windows-task-restart.test.ts b/src/infra/windows-task-restart.test.ts
@@ -120,7 +120,13 @@ describe("relaunchGatewayScheduledTask", () => {
     expect(script).toContain(
       'openclaw restart attempt source=windows-task-handoff target="OpenClaw Gateway (work)"',
     );
+    expect(script).toContain(
+      `powershell.exe -NoProfile -NonInteractive -ExecutionPolicy Bypass -Command "(Get-ScheduledTask -TaskName 'OpenClaw Gateway (work)' -ErrorAction SilentlyContinue).State" 2>nul | findstr /I /C:"Running" >nul 2>&1`,
+    );
     expect(script).toContain('schtasks /Run /TN "OpenClaw Gateway (work)" >>');
+    expect(script.indexOf("powershell.exe -NoProfile")).toBeLessThan(
+      script.indexOf('schtasks /Run /TN "OpenClaw Gateway (work)"'),
+    );
     expect(script).toContain('del "%~f0" >nul 2>&1');
   });
 
@@ -140,6 +146,23 @@ describe("relaunchGatewayScheduledTask", () => {
     expect(script).toContain('schtasks /Run /TN "OpenClaw Gateway (custom)" >>');
   });
 
+  it("escapes custom task names in the PowerShell running-task probe", () => {
+    spawnMock.mockImplementation((_file: string, args: string[]) => {
+      createdScriptPaths.add(decodeCmdPathArg(args[3]));
+      return { unref: vi.fn() };
+    });
+
+    relaunchGatewayScheduledTask({
+      OPENCLAW_WINDOWS_TASK_NAME: "OpenClaw Gateway (Bob's work)",
+    });
+
+    const scriptPath = [...createdScriptPaths][0];
+    const script = fs.readFileSync(scriptPath, "utf8");
+    expect(script).toContain(
+      "-Command \"(Get-ScheduledTask -TaskName 'OpenClaw Gateway (Bob''s work)' -ErrorAction SilentlyContinue).State\"",
+    );
+  });
+
   it("returns failed when the helper cannot be spawned", () => {
     spawnMock.mockImplementation(() => {
       throw new Error("spawn failed");
diff --git a/src/infra/windows-task-restart.ts b/src/infra/windows-task-restart.ts
@@ -13,6 +13,10 @@ import { resolvePreferredOpenClawTmpDir } from "./tmp-openclaw-dir.js";
 const TASK_RESTART_RETRY_LIMIT = 12;
 const TASK_RESTART_RETRY_DELAY_SEC = 1;
 
+function quotePowerShellSingleQuotedLiteral(value: string): string {
+  return `'${value.replace(/'/g, "''")}'`;
+}
+
 function resolveWindowsTaskName(env: NodeJS.ProcessEnv): string {
   const override = env.OPENCLAW_WINDOWS_TASK_NAME?.trim();
   if (override) {
@@ -29,6 +33,10 @@ function buildScheduledTaskRestartScript(params: {
 }): string {
   const { quotedLogPath, setupLines, taskName, taskScriptPath } = params;
   const quotedTaskName = quoteCmdScriptArg(taskName);
+  const queryTaskStateCommand = `(Get-ScheduledTask -TaskName ${quotePowerShellSingleQuotedLiteral(
+    taskName,
+  )} -ErrorAction SilentlyContinue).State`;
+  const quotedQueryTaskStateCommand = quoteCmdScriptArg(queryTaskStateCommand);
   const lines = [
     "@echo off",
     "setlocal",
@@ -40,6 +48,9 @@ function buildScheduledTaskRestartScript(params: {
     ":retry",
     `timeout /t ${TASK_RESTART_RETRY_DELAY_SEC} /nobreak >nul`,
     "set /a attempts+=1",
+    // Avoid racing with another restart path that already started the scheduled task.
+    `powershell.exe -NoProfile -NonInteractive -ExecutionPolicy Bypass -Command ${quotedQueryTaskStateCommand} 2>nul | findstr /I /C:"Running" >nul 2>&1`,
+    "if not errorlevel 1 goto cleanup",
     `schtasks /Run /TN ${quotedTaskName} >> ${quotedLogPath} 2>&1`,
     "if not errorlevel 1 goto cleanup",
     `if %attempts% GEQ ${TASK_RESTART_RETRY_LIMIT} goto fallback`,
