Merge 95939bbea5 into 73f8f3c5fa

back/app.ts

@@ -53,8 +53,19 @@ class Application {
   }
 
   private startMasterProcess() {
+    // Fork gRPC worker first and wait for it to be ready
+    const grpcWorker = this.forkWorker('grpc');
+    
+    // Wait for gRPC worker to signal it's ready before starting HTTP worker
+    this.waitForWorkerReady(grpcWorker, 30000)
+      .then(() => {
+        Logger.info('gRPC worker is ready, starting HTTP worker');
         this.forkWorker('http');
-    this.forkWorker('grpc');
+      })
+      .catch((error) => {
+        Logger.error('Failed to wait for gRPC worker:', error);
+        process.exit(1);
+      });
 
     cluster.on('exit', (worker, code, signal) => {
       const metadata = this.workerMetadataMap.get(worker.id);
@@ -64,10 +75,22 @@ class Application {
             `${metadata.serviceType} worker ${worker.process.pid} died (${signal || code
             }). Restarting...`,
           );
+          // If gRPC worker died, restart it and wait for it to be ready
+          if (metadata.serviceType === 'grpc') {
+            const newGrpcWorker = this.forkWorker('grpc');
+            this.waitForWorkerReady(newGrpcWorker, 30000)
+              .then(() => {
+                Logger.info('gRPC worker restarted and ready');
+              })
+              .catch((error) => {
+                Logger.error('Failed to restart gRPC worker:', error);
+                process.exit(1);
+              });
+          } else {
+            // For HTTP worker, just restart it
             const newWorker = this.forkWorker(metadata.serviceType);
-          Logger.info(
+            Logger.info(`Restarted ${metadata.serviceType} worker (PID: ${newWorker.process.pid})`);
-            `Restarted ${metadata.serviceType} worker (New PID: ${newWorker.process.pid})`,
+          }
-          );
         }
 
         this.workerMetadataMap.delete(worker.id);
@@ -77,6 +100,25 @@ class Application {
     this.setupMasterShutdown();
   }
 
+  private waitForWorkerReady(worker: Worker, timeoutMs: number): Promise<void> {
+    return new Promise<void>((resolve, reject) => {
+      const messageHandler = (msg: any) => {
+        if (msg === 'ready') {
+          worker.removeListener('message', messageHandler);
+          clearTimeout(timeoutId);
+          resolve();
+        }
+      };
+      worker.on('message', messageHandler);
+      
+      // Timeout after specified milliseconds
+      const timeoutId = setTimeout(() => {
+        worker.removeListener('message', messageHandler);
+        reject(new Error(`Worker failed to start within ${timeoutMs / 1000} seconds`));
+      }, timeoutMs);
+    });
+  }
+
   private forkWorker(serviceType: string): Worker {
     const worker = cluster.fork({ SERVICE_TYPE: serviceType });