fix(processmanager): Improve process lifecycle handling and cleanup in daemon, monitors and wrappers

changelog.md

@@ -1,5 +1,13 @@
 # Changelog
 
+## 2025-08-31 - 5.6.2 - fix(processmanager)
+Improve process lifecycle handling and cleanup in daemon, monitors and wrappers
+
+- StartAll: when a monitor exists but is not running, restart it instead of skipping — ensures saved processes are reliably brought online.
+- ProcessMonitor.stop: cancel any pending restart timers to prevent stray restarts after explicit stop.
+- ProcessWrapper: add killProcessTree helper and use it for graceful (SIGTERM) and force (SIGKILL) shutdowns to reliably signal child processes.
+- Daemon stopAll: yield briefly after stopping processes and inspect monitors (not only processInfo) to accurately report stopped vs failed processes.
+
 ## 2025-08-31 - 5.6.1 - fix(daemon)
 Ensure robust process shutdown and improve logs/subscriber diagnostics
 

ts/00_commitinfo_data.ts

@@ -3,6 +3,6 @@
  */
 export const commitinfo = {
   name: '@git.zone/tspm',
-  version: '5.6.1',
+  version: '5.6.2',
   description: 'a no fuzz process manager'
 }

ts/daemon/processmanager.ts

@@ -499,8 +499,12 @@ export class ProcessManager extends EventEmitter {
    */
   public async startAll(): Promise<void> {
     for (const [id, config] of this.processConfigs.entries()) {
-      if (!this.processes.has(id)) {
+      const monitor = this.processes.get(id);
+      if (!monitor) {
         await this.start(config);
+      } else if (!monitor.isRunning()) {
+        // If a monitor exists but is not running, restart the process to ensure a clean start
+        await this.restart(id);
       }
     }
   }

ts/daemon/processmonitor.ts

@@ -400,6 +400,11 @@ export class ProcessMonitor extends EventEmitter {
     if (this.intervalId) {
       clearInterval(this.intervalId);
     }
+    // Cancel any pending restart timer
+    if (this.restartTimer) {
+      clearTimeout(this.restartTimer);
+      this.restartTimer = null;
+    }
     if (this.processWrapper) {
       // Clear pidusage state for current PID before stopping to avoid leaks
       try {

ts/daemon/processwrapper.ts

@@ -23,6 +23,26 @@ export class ProcessWrapper extends EventEmitter {
   private runId: string = '';
   private stdoutRemainder: string = '';
   private stderrRemainder: string = '';
+  
+  // Helper: send a signal to the process and all its children (best-effort)
+  private async killProcessTree(signal: NodeJS.Signals): Promise<void> {
+    if (!this.process || !this.process.pid) return;
+    const rootPid = this.process.pid;
+    await new Promise<void>((resolve) => {
+      plugins.psTree(rootPid, (err: any, children: ReadonlyArray<{ PID: string }>) => {
+        const pids: number[] = [rootPid, ...children.map((c) => Number(c.PID)).filter((n) => Number.isFinite(n))];
+        for (const pid of pids) {
+          try {
+            // Always signal individual PIDs to avoid accidentally targeting unrelated groups
+            process.kill(pid, signal);
+          } catch {
+            // ignore ESRCH/EPERM
+          }
+        }
+        resolve();
+      });
+    });
+  }
 
   constructor(options: IProcessWrapperOptions) {
     super();
@@ -193,17 +213,13 @@ export class ProcessWrapper extends EventEmitter {
     // First try SIGTERM for graceful shutdown
     if (this.process.pid) {
       try {
-        this.logger.debug(`Sending SIGTERM to process ${this.process.pid}`);
+        this.logger.debug(`Sending SIGTERM to process tree rooted at ${this.process.pid}`);
-        try {
+        await this.killProcessTree('SIGTERM');
-          // Try to signal the whole process group on POSIX to ensure children get the signal too
+
-          if (process.platform !== 'win32') {
+        // If the process already exited, return immediately
-            process.kill(-Math.abs(this.process.pid), 'SIGTERM');
+        if (typeof this.process.exitCode === 'number') {
-          } else {
+          this.logger.debug('Process already exited, no need to wait');
-            process.kill(this.process.pid, 'SIGTERM');
+          return;
-          }
-        } catch {
-          // Fallback to direct process kill if group kill fails
-          process.kill(this.process.pid, 'SIGTERM');
         }
 
         // Wait for exit or escalate
@@ -218,26 +234,15 @@ export class ProcessWrapper extends EventEmitter {
           const onExit = () => cleanup();
           this.process!.once('exit', onExit);
 
-          const killTimer = setTimeout(() => {
+          const killTimer = setTimeout(async () => {
             if (!this.process || !this.process.pid) return cleanup();
             this.logger.warn(
-              `Process ${this.process.pid} did not exit gracefully, force killing...`,
+              `Process ${this.process.pid} did not exit gracefully, force killing tree...`,
-            );
-            this.addSystemLog(
-              'Process did not exit gracefully, force killing...',
             );
+            this.addSystemLog('Process did not exit gracefully, force killing...');
             try {
-              if (process.platform !== 'win32') {
+              await this.killProcessTree('SIGKILL');
-                process.kill(-Math.abs(this.process.pid), 'SIGKILL');
+            } catch {}
-              } else {
-                process.kill(this.process.pid, 'SIGKILL');
-              }
-            } catch (error: any) {
-              this.logger.debug(
-                `Failed to send SIGKILL, process probably already exited: ${error?.message || String(error)}`,
-              );
-            }
-
             // Give a short grace period after SIGKILL
             setTimeout(() => cleanup(), 500);
           }, 5000);

ts/daemon/tspm.daemon.ts

@@ -450,10 +450,12 @@ export class TspmDaemon {
 
         await this.tspmInstance.setDesiredStateForAll('stopped');
         await this.tspmInstance.stopAll();
+        // Yield briefly to allow any pending exit events to settle
+        await new Promise((r) => setTimeout(r, 50));
 
-        // Get status of all processes
+        // Determine which monitors are no longer running
-        for (const [id, processInfo] of this.tspmInstance.processInfo) {
+        for (const [id, monitor] of this.tspmInstance.processes) {
-          if (processInfo.status === 'stopped') {
+          if (!monitor.isRunning()) {
             stopped.push(id);
           } else {
             failed.push({ id, error: 'Failed to stop' });