@shepai/cli 1.130.0-pr413.0773bb9 → 1.130.0-pr413.cb198fb

package/dist/packages/core/src/infrastructure/services/deployment/deployment.service.d.ts

@@ -31,7 +31,16 @@ export declare class DeploymentService implements IDeploymentService {
     setDatabase(db: Database.Database): void;
     /**
      * Recover dev servers from the database on startup.
-     * Validates each PID is still alive; removes dead rows.
+     *
+     * Three recovery strategies based on process state:
+     * 1. Alive + Ready (has URL) → re-adopt, then validate URL asynchronously
+     * 2. Alive + Booting (no URL) → kill orphan and re-spawn (can't re-attach stdout)
+     * 3. Dead → re-spawn if target directory still exists
+     *
+     * After sync recovery, fires off async URL health checks for re-adopted
+     * processes. Zombies (alive but not serving) are killed and re-spawned.
+     *
+     * All re-spawns are wrapped in try/catch so one failure doesn't block others.
      */
     recoverAll(): void;
     /**
@@ -77,5 +86,16 @@ export declare class DeploymentService implements IDeploymentService {
     private attachOutputListener;
     private pollUntilDead;
     private waitForExit;
+    /**
+     * Async health check for a recovered entry. Probes the URL port via TCP.
+     * If the port is not responding (zombie process), kills and re-spawns.
+     * Fire-and-forget — errors are logged, never thrown.
+     */
+    private validateAndRespawn;
+    /**
+     * TCP connect probe to check if a URL's port is accepting connections.
+     * Returns true if the port responds within 2 seconds, false otherwise.
+     */
+    private probePort;
 }
 //# sourceMappingURL=deployment.service.d.ts.map

package/dist/packages/core/src/infrastructure/services/deployment/deployment.service.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"deployment.service.d.ts","sourceRoot":"","sources":["../../../../../../../packages/core/src/infrastructure/services/deployment/deployment.service.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;GAUG;AAEH,OAAO,EAAE,KAAK,EAAmC,MAAM,oBAAoB,CAAC;AAE5E,OAAO,KAAK,QAAQ,MAAM,gBAAgB,CAAC;AAK3C,OAAO,KAAK,EACV,kBAAkB,EAClB,gBAAgB,EAChB,QAAQ,EACT,MAAM,qEAAqE,CAAC;AAC7E,OAAO,EAAE,eAAe,EAAE,MAAM,wBAAwB,CAAC;AAiCzD,MAAM,WAAW,qBAAqB;IACpC,KAAK,EAAE,OAAO,KAAK,CAAC;IACpB,eAAe,EAAE,OAAO,eAAe,CAAC;IACxC,IAAI,EAAE,CAAC,GAAG,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,CAAC,OAAO,GAAG,MAAM,KAAK,IAAI,CAAC;IAC7D,OAAO,EAAE,CAAC,GAAG,EAAE,MAAM,KAAK,OAAO,CAAC;CACnC;AA0CD,qBAAa,iBAAkB,YAAW,kBAAkB;IAC1D,OAAO,CAAC,QAAQ,CAAC,WAAW,CAAsC;IAClE,OAAO,CAAC,QAAQ,CAAC,IAAI,CAAwB;IAC7C,OAAO,CAAC,QAAQ,CAAC,OAAO,CAAsB;IAC9C,OAAO,CAAC,EAAE,CAAkC;gBAEhC,IAAI,GAAE,OAAO,CAAC,qBAAqB,CAAM;IAIrD;;OAEG;IACH,WAAW,CAAC,EAAE,EAAE,QAAQ,CAAC,QAAQ,GAAG,IAAI;IAIxC;;;OAGG;IACH,UAAU,IAAI,IAAI;IA4ClB;;;OAGG;IACH,KAAK,CAAC,QAAQ,EAAE,MAAM,EAAE,UAAU,EAAE,MAAM,EAAE,UAAU,SAAe,GAAG,IAAI;IA0G5E;;;;OAIG;IACH,SAAS,CAAC,QAAQ,EAAE,MAAM,GAAG,gBAAgB,GAAG,IAAI;IA4CpD;;OAEG;IACG,IAAI,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC,IAAI,CAAC;IA0D3C;;OAEG;IACH,OAAO,IAAI,IAAI;IAef;;OAEG;IACH,OAAO,CAAC,QAAQ,EAAE,MAAM,GAAG,QAAQ,EAAE,GAAG,IAAI;IAM5C;;OAEG;IACH,EAAE,CAAC,KAAK,EAAE,KAAK,EAAE,OAAO,EAAE,CAAC,KAAK,EAAE,QAAQ,KAAK,IAAI,GAAG,IAAI;IAI1D;;OAEG;IACH,GAAG,CAAC,KAAK,EAAE,KAAK,EAAE,OAAO,EAAE,CAAC,KAAK,EAAE,QAAQ,KAAK,IAAI,GAAG,IAAI;IAM3D,OAAO,CAAC,QAAQ;IAuBhB,OAAO,CAAC,aAAa;IAWrB,OAAO,CAAC,QAAQ;IAShB,OAAO,CAAC,MAAM;IAed,OAAO,CAAC,WAAW;IAQnB;;;OAGG;IACH,OAAO,CAAC,oBAAoB;YA4Dd,aAAa;IAW3B,OAAO,CAAC,WAAW;CASpB"}
+{"version":3,"file":"deployment.service.d.ts","sourceRoot":"","sources":["../../../../../../../packages/core/src/infrastructure/services/deployment/deployment.service.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;GAUG;AAEH,OAAO,EAAE,KAAK,EAAmC,MAAM,oBAAoB,CAAC;AAK5E,OAAO,KAAK,QAAQ,MAAM,gBAAgB,CAAC;AAK3C,OAAO,KAAK,EACV,kBAAkB,EAClB,gBAAgB,EAChB,QAAQ,EACT,MAAM,qEAAqE,CAAC;AAC7E,OAAO,EAAE,eAAe,EAAE,MAAM,wBAAwB,CAAC;AAiCzD,MAAM,WAAW,qBAAqB;IACpC,KAAK,EAAE,OAAO,KAAK,CAAC;IACpB,eAAe,EAAE,OAAO,eAAe,CAAC;IACxC,IAAI,EAAE,CAAC,GAAG,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,CAAC,OAAO,GAAG,MAAM,KAAK,IAAI,CAAC;IAC7D,OAAO,EAAE,CAAC,GAAG,EAAE,MAAM,KAAK,OAAO,CAAC;CACnC;AA0CD,qBAAa,iBAAkB,YAAW,kBAAkB;IAC1D,OAAO,CAAC,QAAQ,CAAC,WAAW,CAAsC;IAClE,OAAO,CAAC,QAAQ,CAAC,IAAI,CAAwB;IAC7C,OAAO,CAAC,QAAQ,CAAC,OAAO,CAAsB;IAC9C,OAAO,CAAC,EAAE,CAAkC;gBAEhC,IAAI,GAAE,OAAO,CAAC,qBAAqB,CAAM;IAIrD;;OAEG;IACH,WAAW,CAAC,EAAE,EAAE,QAAQ,CAAC,QAAQ,GAAG,IAAI;IAIxC;;;;;;;;;;;;OAYG;IACH,UAAU,IAAI,IAAI;IAgFlB;;;OAGG;IACH,KAAK,CAAC,QAAQ,EAAE,MAAM,EAAE,UAAU,EAAE,MAAM,EAAE,UAAU,SAAe,GAAG,IAAI;IA0G5E;;;;OAIG;IACH,SAAS,CAAC,QAAQ,EAAE,MAAM,GAAG,gBAAgB,GAAG,IAAI;IA4CpD;;OAEG;IACG,IAAI,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC,IAAI,CAAC;IA0D3C;;OAEG;IACH,OAAO,IAAI,IAAI;IAef;;OAEG;IACH,OAAO,CAAC,QAAQ,EAAE,MAAM,GAAG,QAAQ,EAAE,GAAG,IAAI;IAM5C;;OAEG;IACH,EAAE,CAAC,KAAK,EAAE,KAAK,EAAE,OAAO,EAAE,CAAC,KAAK,EAAE,QAAQ,KAAK,IAAI,GAAG,IAAI;IAI1D;;OAEG;IACH,GAAG,CAAC,KAAK,EAAE,KAAK,EAAE,OAAO,EAAE,CAAC,KAAK,EAAE,QAAQ,KAAK,IAAI,GAAG,IAAI;IAM3D,OAAO,CAAC,QAAQ;IAuBhB,OAAO,CAAC,aAAa;IAWrB,OAAO,CAAC,QAAQ;IAShB,OAAO,CAAC,MAAM;IAed,OAAO,CAAC,WAAW;IAQnB;;;OAGG;IACH,OAAO,CAAC,oBAAoB;YA4Dd,aAAa;IAW3B,OAAO,CAAC,WAAW;IAUnB;;;;OAIG;YACW,kBAAkB;IAsChC;;;OAGG;IACH,OAAO,CAAC,SAAS;CAwBlB"}

package/dist/packages/core/src/infrastructure/services/deployment/deployment.service.js

@@ -10,6 +10,9 @@
  * process state (clean up dead PIDs, re-adopt live ones).
  */
 import { spawn, execFileSync } from 'node:child_process';
+import { existsSync } from 'node:fs';
+import { join } from 'node:path';
+import net from 'node:net';
 import { EventEmitter } from 'node:events';
 // On Unix we use process.kill(-pid) to send signals to the process GROUP
 // (detached: true puts the child in its own group via setsid()).
@@ -82,7 +85,16 @@ export class DeploymentService {
     }
     /**
      * Recover dev servers from the database on startup.
-     * Validates each PID is still alive; removes dead rows.
+     *
+     * Three recovery strategies based on process state:
+     * 1. Alive + Ready (has URL) → re-adopt, then validate URL asynchronously
+     * 2. Alive + Booting (no URL) → kill orphan and re-spawn (can't re-attach stdout)
+     * 3. Dead → re-spawn if target directory still exists
+     *
+     * After sync recovery, fires off async URL health checks for re-adopted
+     * processes. Zombies (alive but not serving) are killed and re-spawned.
+     *
+     * All re-spawns are wrapped in try/catch so one failure doesn't block others.
      */
     recoverAll() {
         if (!this.db)
@@ -102,7 +114,10 @@ export class DeploymentService {
         }
         log.info(`Recovering ${rows.length} dev server(s) from database`);
         for (const row of rows) {
-            if (this.deps.isAlive(row.pid)) {
+            const alive = this.deps.isAlive(row.pid);
+            const hasUrl = row.state === DeploymentState.Ready && row.url;
+            if (alive && hasUrl) {
+                // Strategy 1: Process alive and Ready — re-adopt, then validate URL async
                 log.info(`Recovered "${row.target_id}" (pid=${row.pid}, state=${row.state}, url=${row.url})`);
                 const entry = {
                     pid: row.pid,
@@ -117,10 +132,36 @@ export class DeploymentService {
                     logs: new LogRingBuffer(),
                 };
                 this.deployments.set(row.target_id, entry);
+                // Fire-and-forget: validate URL is actually responding, kill zombie if not
+                this.validateAndRespawn(entry);
+                continue;
+            }
+            if (alive) {
+                // Strategy 2: Process alive but stuck in Booting — can't re-attach stdout
+                // Kill the orphan so we can re-spawn with proper listeners
+                log.info(`Orphan "${row.target_id}" (pid=${row.pid}) stuck in ${row.state} — killing to re-spawn`);
+                try {
+                    this.deps.kill(row.pid, 'SIGKILL');
+                }
+                catch {
+                    // already dead
+                }
             }
             else {
-                log.info(`Stale dev server "${row.target_id}" (pid=${row.pid}) — removing from DB`);
-                this.dbDelete(row.target_id);
+                log.info(`Dev server "${row.target_id}" (pid=${row.pid}) is dead — will re-spawn`);
+            }
+            // Strategy 2 & 3: Re-spawn if target directory still has a package.json
+            this.dbDelete(row.target_id);
+            if (!existsSync(join(row.target_path, 'package.json'))) {
+                log.warn(`Skipping re-spawn for "${row.target_id}" — no package.json at "${row.target_path}"`);
+                continue;
+            }
+            try {
+                log.info(`Re-spawning dev server for "${row.target_id}" at "${row.target_path}"`);
+                this.start(row.target_id, row.target_path, row.target_type);
+            }
+            catch (err) {
+                log.error(`Failed to re-spawn "${row.target_id}": ${err}`);
             }
         }
     }
@@ -486,4 +527,69 @@ export class DeploymentService {
             });
         });
     }
+    /**
+     * Async health check for a recovered entry. Probes the URL port via TCP.
+     * If the port is not responding (zombie process), kills and re-spawns.
+     * Fire-and-forget — errors are logged, never thrown.
+     */
+    async validateAndRespawn(entry) {
+        if (!entry.url)
+            return;
+        const responding = await this.probePort(entry.url);
+        if (responding) {
+            log.info(`[${entry.targetId}] URL health check passed — "${entry.url}" is responding`);
+            return;
+        }
+        log.warn(`[${entry.targetId}] URL health check FAILED — "${entry.url}" not responding, killing zombie (pid=${entry.pid})`);
+        // Kill the zombie
+        try {
+            this.deps.kill(entry.pid, 'SIGKILL');
+        }
+        catch {
+            // already dead
+        }
+        this.deployments.delete(entry.targetId);
+        this.dbDelete(entry.targetId);
+        // Re-spawn if target still exists
+        if (!existsSync(join(entry.targetPath, 'package.json'))) {
+            log.warn(`Skipping re-spawn for "${entry.targetId}" — no package.json at "${entry.targetPath}"`);
+            return;
+        }
+        try {
+            log.info(`Re-spawning dev server for "${entry.targetId}" at "${entry.targetPath}"`);
+            this.start(entry.targetId, entry.targetPath, entry.targetType);
+        }
+        catch (err) {
+            log.error(`Failed to re-spawn "${entry.targetId}": ${err}`);
+        }
+    }
+    /**
+     * TCP connect probe to check if a URL's port is accepting connections.
+     * Returns true if the port responds within 2 seconds, false otherwise.
+     */
+    probePort(url) {
+        return new Promise((resolve) => {
+            try {
+                const parsed = new URL(url);
+                const port = parseInt(parsed.port || (parsed.protocol === 'https:' ? '443' : '80'), 10);
+                const socket = net.createConnection({ host: parsed.hostname, port });
+                const timeout = setTimeout(() => {
+                    socket.destroy();
+                    resolve(false);
+                }, 2000);
+                socket.on('connect', () => {
+                    clearTimeout(timeout);
+                    socket.destroy();
+                    resolve(true);
+                });
+                socket.on('error', () => {
+                    clearTimeout(timeout);
+                    resolve(false);
+                });
+            }
+            catch {
+                resolve(false);
+            }
+        });
+    }
 }

package/dist/src/presentation/web/dev-server.js

@@ -107,6 +107,13 @@ async function main() {
         console.log('\n[dev-server] Shutting down...');
         const forceExit = setTimeout(() => process.exit(0), 2000);
         try {
+            try {
+                const deploymentService = container.resolve('IDeploymentService');
+                deploymentService.stopAll();
+            }
+            catch {
+                /* not initialized */
+            }
             try {
                 getNotificationWatcher().stop();
             }