[QUESTION] Stop bot with infinite background job/task gracefully

[max-dw-i]

Issue I am facing

I have a bot with an infinite background job. What is the best, "canonical" way of stopping/quitting the bot in this case (with saving the state to a persistent storage and other stuff that Application.stop() does). Ideally, it should work not only when I press ctrl-c but in any scenario when the bot stops (for example, when an exception that could not be handled for some reason was raised and the bot stopped).
So If I press ctrl-c, the bot does not stop. I need to press ctrl-c one more time and the process does not finish properly (e.g. 'before stop' persistence does not work).

Traceback to the issue

No response

Related part of your code

class Bot:
    def init(self, app: Application) -> None:
        self.app = app
        self.app.job_queue.run_once(
            callback=self.infinite_task,
            when=0
        )

    async def infinite_task(self) -> None:
        while True:
            print("I'm a bot")
            await asyncio.sleep(1)

application = ApplicationBuilder().token('TOKEN').build()
Bot(application)
application.run_polling()

Operating System

Debian 10

Version of Python, python-telegram-bot & dependencies

python-telegram-bot 20.0a2
Bot API 6.1
Python 3.7.3 (default, Jan 22 2021, 20:04:44)  [GCC 8.3.0]

[Bibo-Joshi]

As a first step I would try to get rid of the infinite loop. Those should be reserved to very few critical tasks like polling for updates or so. Alternatives can include e.g. using JobQueue.run_repeating.
If it is definitely not possible to avoid an inifinite loop, then I would recommend to decouple it from any PTB-logic as much as possible. Using the run_once to execute the loop is an abuse of JobQueue IMO. A cleaner approach would be something along the lines

infinite_tasks = []

async def post_init(app):
    infinite_tasks.append(asyncio.create_task(infinite_task()))


...
app.run_polling(close_loop=False)
for task in infinite_tasks:
    task.cancel()

(Haven't tested the code) Or use asyncio.Event objects to signal to the loops that they should stop.
Note that in the next version, there will also be a method Application.post_shutdown which will make logic like this even simpler.

[max-dw-i]

1. Is there any way to check if bot is shutting down? Flag that is set True when the process starts or something? Or pre_shutdown hook that is called before anything involved in shutting down runs? Could not find anything in the docs.
2. What happens if a unhandled (for some reason) exception is raised and a bot is stopped? Is update_persistence called?

[Bibo-Joshi]

@oratosquilla-oratoria there is telegram.ext.Application.running. If you would explain your use case, one could probably help you better 😉