fix(core): missing check for double removal of unresponsive module (#850

)

* fix(core): missing check for double removal of unresponsive module
refactor(core): add sanity check for module existence in monitor loop

* chore: add more logs in module recovery

packages/core/src/config-manager/service-discovery/ServiceMonitor.ts

@@ -169,15 +169,18 @@ export class ServiceMonitor {
 
   private async monitorModules() {
     for (const module of this._serviceRegistry.getRegisteredModules()) {
-      const registeredModule = this._serviceRegistry.getModule(module)!;
+      const registeredModule = this._serviceRegistry.getModule(module);
+      if (!registeredModule) continue;
       try {
         await this.healthCheckService(module, registeredModule.address);
       } catch (e) {
-        this.handleUnresponsiveModule(
-          module,
-          registeredModule.address,
-          HealthCheckStatus.SERVICE_UNKNOWN,
-        );
+        if (this._serviceRegistry.getModule(module)) {
+          this.handleUnresponsiveModule(
+            module,
+            registeredModule.address,
+            HealthCheckStatus.SERVICE_UNKNOWN,
+          );
+        }
       }
     }
     this.moduleRegister.emit('serving-modules-update');

packages/core/src/config-manager/service-discovery/index.ts

@@ -137,6 +137,8 @@ export class ServiceDiscovery {
         this.grpcSdk.createModuleClient(moduleName, moduleUrl);
       }
     } catch (e) {
+      ConduitGrpcSdk.Logger.error(`SD: failed to recover: ${moduleName} ${moduleUrl}`);
+      ConduitGrpcSdk.Logger.error(`SD: recovery error: ${e}`);
       throw new Error('Failed to register unresponsive module');
     }
     const healthStatus = healthResponse.status as unknown as HealthCheckStatus;