Adding BullModule.forRootAsync to CliModule makes it hang forever

[simplenotezy]

If I add:

BullModule.forRootAsync({
  useFactory: ConfigFactory.bullConfig,
  inject: [ConfigService],
}),

To my CliModule it will always run, so if we have cron jobs that uses a cli command it will keep reporting as 'Running'.

If I remove the BullModule the command will exit as usual.

[jmcdo29]

Interesting, I would expect the static run to end up calling await app.close() after the command runner service's run method finished:

nest-commander/packages/nest-commander/src/command.factory.ts

Lines 16 to 23 in 7ed3feb

	static async run(
	rootModule: Type<any>,
	optionsOrLogger?: CommandFactoryRunOptions | NestLogger,
	): Promise<void> {
	const app = await this.createWithoutRunning(rootModule, optionsOrLogger);
	await this.runApplication(app);
	await app.close();
	}

nest-commander/packages/nest-commander/src/command.factory.ts

Lines 56 to 62 in 7ed3feb

	static async runApplication(
	app: INestApplicationContext,
	): Promise<INestApplicationContext> {
	const runner = app.get(CommandRunnerService);
	await runner.run();
	return app;
	}

[jmcdo29]

Thanks for the reproduction. I'm pretty certain this is an issue with the @nestjs/bull library, as it doesn't implement an OnModuleClose to handle closing the connection to bull or redis when the application shuts down. nest-commander is properly closing the NestJS application, but if that connection pool is still open, then the Node event loop never resolves and the process is never finished

[simplenotezy]

I see, thanks @jmcdo29 - I have created an issue for bull here: nestjs/bull#1705

[jmcdo29]

Interesting. So Nest does have those methods (hard to find, but they are there). Digging in a bit more, I do see that the Queue's close is indeed called, so now it's a question of why the delay. It looks like after half a minute the application is properly closed (see screenshot), so it's likely just redis being slow to kill the connection. Very annoying to say the least

[simplenotezy]

Hmm, interesting. Well, in my case it hangs for hours. If I look at my CLI I see it hanging if I forgot to press CTRL+C even after hours.

Can you think of a workaround? Redis closes down fast in production (outside CLI) on graceful shutdowns etc, so it's indeed odd.

So far I am calling exit(0), but that causes a non-graceful-shutdown unfortunately.

[simplenotezy]

Created a ticket for it here: #1042