Sending SIGTERM to parent process when running with --workers hangs indefinitely

Checklist

• The bug is reproducible against the latest release and/or master.
• There are no similar issues or pull requests to fix it yet.

Describe the bug

When running uvicorn with multiple workers, sending a SIGTERM ie kill -15 ppid there’s no graceful shutdown and the process hangs indefinitely on https://github.com/encode/uvicorn/blob/ff4af12d6902bc9d535fe2a948d1df3ffa02b0d3/uvicorn/supervisors/multiprocess.py#L57

To reproduce

Expected behavior

A graceful shutdown

Actual behavior

Debugging material

Environment

• OS / Python / Uvicorn version: just run uvicorn --version Running uvicorn 0.12.2 with CPython 3.8.6 on Linux
• The exact command you’re running uvicorn with, all flags you passed included. If you run it with gunicorn please do the same. If there is a reverse-proxy involved and you cannot reproduce without it please give the minimal config of it to reproduce.

Additional context

originally discovered in https://github.com/encode/uvicorn/issues/364 and mistakenly taken for a docker issue, it’s in fact a SIGTERM issue afaiu

[euri10]

yep we got a problem with SIGTERM, and not just in docker: it was discovered 1st as a docker issue because it’s the default signal sent to exit a container , but the issue exists without containers, so the digging performed in #364 that ended up in a potential fix in #620 was not the right move.

I think #853 is the correct way to handle multi process exits cleanly, it is just too big to review, cant find a way not time to slice it properly

[sgsabbage]

As far as I can tell, calling terminate() just sends a SIGTERM to each of the child processes, which then handle it appropriately. As it stands the parent process eats the SIGTERM and so the children never see it and don’t know to terminate.

PR inbound!