cannot pickle _io.TextiOWrapper about granian

[gi0baro]

Looks like you're passing your application instance as the app parameter, while that argument should be a string referring to where to import your application from.

[albertoreis86]

not really actually,
this is how I instantiate:

granian = Granian(target=f'{settings.APPLICATION_PATH}',
address=settings.SERVER_HOST,
port=settings.PORT,
interface=constants.Interfaces.ASGI,
workers=settings.WORKERS_NUMBER,
log_level=log.LogLevels.debug if settings.DEBUG else log.LogLevels.info,
ssl_cert=cert_file_path if cert_file_path.exists() else None,
ssl_key=cert_keyfile_path if cert_keyfile_path.exists() else None,
reload=settings.RELOAD_SERVER,
log_dictconfig=logging_config)
granian.serve()

[gi0baro]

That's strange 🤔

What kind of application is that? Have you ever successfully run it with another server and workers > 1?
Also it sounds confusing to me the stacktrace looks like it's macOs, but instead of using spawn as the multiprocessing context (which is the default under macOs) it uses fork.

[albertoreis86]

Yes we use to run this with hypercorn and was going ok.,
indeed it is macOS, the other members of the team can run it normally using linux.
the default value for workers number is 1.

[gi0baro]

Ok so probably if you call:

import multiprocessing
multiprocessing.set_start_method('fork')

you should be able to run it in the same way it works on Linux.

What it still sound strange to me, is why you have an _io.TextIOWrapper instance defined somewhere before starting granian. It sounds like you opened a file or a stream and never closed it, or never parsed the context of that stream. Probably just a str(variable) would make the thing pickable.

[albertoreis86]

Thanks @gi0baro it worked explicitly setting the start method to fork.
I will dig into the _io.TextIOWrapper I will dig into it.
Thanks