[Core] ray.init() hangs using Python 3.10.15 on Linux

[btakita]

What happened + What you expected to happen

1. ray.init() hangs
2. ray.init() should start & not hang

ray-logs.zip

Versions / Dependencies

Ray 2.38.0
Python 3.10.15
OS Linux 6.11.6-arch1-1 x86_64 unknown

Reproduction script

import ray.util.queue
ray.init()

Issue Severity

High: It blocks me from completing my task.

[btakita]

I have also noticed some processes that I have not figured out how to kill -9.

✗ ps aux | grep python
brian    3197578  0.0  0.1 1932304 105808 pts/10 D    06:52   0:00 python3.10 -m run start
brian    3197660  0.0  0.0      0     0 pts/10   Z    06:52   0:00 [python3.10] <defunct>
brian    3197661  0.0  0.0      0     0 pts/10   Z    06:52   0:00 [python3.10] <defunct>
brian    3202122  0.0  0.1 1932312 105704 pts/10 D    06:53   0:00 python3.10 -m run start
brian    3202205  0.0  0.0      0     0 pts/10   Z    06:53   0:00 [python3.10] <defunct>
brian    3202206  0.0  0.0      0     0 pts/10   Z    06:53   0:00 [python3.10] <defunct>
brian    3212473  0.0  0.1 1932304 105684 pts/10 D    06:55   0:00 python3.10 -m run start
brian    3212553  0.0  0.0      0     0 pts/10   Z    06:55   0:00 [python3.10] <defunct>
brian    3212554  0.0  0.0      0     0 pts/10   Z    06:55   0:00 [python3.10] <defunct>
brian    3218399  0.0  0.1 1932440 105832 pts/10 D    06:56   0:00 python3.10 -m run start
brian    3218804  0.0  0.0      0     0 pts/10   Z    06:56   0:00 [python3.10] <defunct>
brian    3218819  0.0  0.0      0     0 pts/10   Z    06:56   0:00 [python3.10] <defunct>
brian    3225056  0.0  0.1 1932368 106128 pts/10 D    06:58   0:01 python3.10 -m run start
brian    3225138  0.0  0.0      0     0 pts/10   Z    06:58   0:00 [python3.10] <defunct>
brian    3225139  0.0  0.0      0     0 pts/10   Z    06:58   0:00 [python3.10] <defunct>
brian    3324666  0.0  0.0 1883792 89044 pts/10  D    07:16   0:00 python3.10 -m run start
brian    3324750  0.0  0.0      0     0 pts/10   Z    07:16   0:00 [python3.10] <defunct>
brian    3324751  0.0  0.0      0     0 pts/10   Z    07:16   0:00 [python3.10] <defunct>
brian    3344760  0.0  0.0 1883712 88936 pts/10  D+   07:20   0:00 /home/brian/work/livekit/agent/.venv/bin/python -m ray_test
brian    3345025  0.0  0.0      0     0 pts/10   Z+   07:20   0:00 [python] <defunct>
brian    3345026  0.0  0.0      0     0 pts/10   Z+   07:20   0:00 [python] <defunct>
brian    3411983  0.0  0.0   6396  3704 pts/17   S+   07:32   0:00 grep --color=auto python

[Superskyyy]

Can you try a ray stop --force then redo the init? Or start ray from the command line see if it works.

[btakita]

Thank you for the tip!
I was having power issues on my laptop & my 2nd m.2 drive shut down. I have consolidated my m.2 drives & reinstalled the OS. My laptop seems to be more stable now. If I have this problem independent of the power issues, I'll reopen this GH issue.

[btakita]

A bit more context, the m.2 drive that shut down had the /home directory. All the other root directories in the system, including /tmp were on the m.2 drive that was still running.

I was running BTRFS with the issue & am now running EXT4. I was also having issues with video playback stuttering (buffer running out) & it seems to have gone away with EXT4 on the fresh install.

[btakita]

I reinstalled my OS & don't have the power issues. But ray.init() is hanging again. So I'm reopening. I didn't run ray stop --force. But next time I will.

It seems to happen after I use a stop the driver using a KeyboardInterrupt then start then stop using KeyboardInterrupt a few times. In a dev environment.

[rynewang]

Hi can you try to create a fresh conda environment and make sure there's no ray process by ray stop --force, then do the ray.init() to test? If it hangs, would you mind to give us logs of the python script, and /tmp/ray?

[btakita]

It froze again & I ran ray stop --force.
ray-logs.zip

$: ./.venv/bin/ray stop --force
Stopped only 0 out of 3 Ray processes within the grace period 16 seconds. Set `-v` to see more details. Remaining processes [psutil.Process(pid=3773004, name='python3.10', status='zombie', started='22:39:32'), psutil.Process(pid=3773005, name='python3.10', status='zombie', started='22:39:32'), psutil.Process(pid=3772927, name='gcs_server', status='zombie', started='22:39:32')] will be forcefully terminated.
You can also use `--force` to forcefully terminate processes or set higher `--grace-period` to wait longer time for proper termination.

I reran the process & ray.init() hung again.

[j93hahn]

+1, I'm using python 3.10.15 and it hangs on my mac m4:

❯ uv run ipython
Python 3.10.15 (main, Oct 16 2024, 08:33:15) [Clang 18.1.8 ]
Type 'copyright', 'credits' or 'license' for more information
IPython 8.32.0 -- An enhanced Interactive Python. Type '?' for help.

In [1]: import ray

In [2]: ray.init()
2025-03-04 11:47:34,281	INFO worker.py:1832 -- Started a local Ray instance. View the dashboard at 127.0.0.1:8265
Out[5]: RayContext(dashboard_url='127.0.0.1:8265', python_version='3.10.15', ray_version='2.42.1', ray_commit='c2e38f7b75be223c0c033986472daada8622d64f')

I've run uv run ray stop --force and pkill but nothing seems to address this