# Issues with setting up runtime environment using "uv run"

**URL:** <https://discuss.ray.io/t/issues-with-setting-up-runtime-environment-using-uv-run/23328>\
**Category:** Ray Core\
**Created:** [November 19, 2025, 10:09am UTC](https://discuss.ray.io/t/issues-with-setting-up-runtime-environment-using-uv-run/23328 "2025-11-19T10:09:17Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![bln\_prasad](https://sea2.discourse-cdn.com/flex020/user_avatar/discuss.ray.io/bln_prasad/32/8328_2.png) [@bln\_prasad](https://discuss.ray.io/u/bln_prasad)\
**Post date:** [November 19, 2025, 10:09am UTC](https://discuss.ray.io/t/issues-with-setting-up-runtime-environment-using-uv-run/23328/1 "2025-11-19T10:09:17Z")

</div>

I’m experimenting with running my application job using **UV** , but it’s not working as expected.  
When I submit the job using: **uv python pin 3.9 && uv run -m myapp.mytask** along with the runtime configuration below, the job remains in a **PENDING** state. There are no logs available that indicate what is happening internally.

```auto
runtime_env={
   "env_vars": {    
   "py_executable": "uv python pin 3.9 && uv run", ==> tried with just "uv run"
   "working_dir": "../../nile-intelligence",
   "excludes":[".venv"]
}

```

I tested the setup without using the `"py_executable"` configuration, and it works without any issues. However, since I need to run tasks across nodes, I have to configure the same runtime environment on all nodes. When I use `"py_executable"`, the job hangs, and there are no clear indications of where it gets stuck.

I’m using the **Ray 2.46** image for the cluster. Are there any known issues with this version aside from the inability to stop pending jobs?

Any guidance or suggestions would be appreciated.

---

<div class="post-metadata">

**Author:** ![RunLLM](https://sea2.discourse-cdn.com/flex020/user_avatar/discuss.ray.io/runllm/32/8249_2.png) [@RunLLM](https://discuss.ray.io/u/RunLLM)\
**Post date:** [November 19, 2025, 10:10am UTC](https://discuss.ray.io/t/issues-with-setting-up-runtime-environment-using-uv-run/23328/2 "2025-11-19T10:10:14Z")

</div>

Yes, there are known issues with using `"py_executable": "uv run"` in Ray runtime environments, especially in Ray 2.46. Jobs can hang in the PENDING state with no logs, and stopping such jobs is also problematic. This is a common pain point reported by users, and improvements/fixes have been made in later Ray versions ([see issue](https://github.com/ray-project/ray/issues/56166)). Additionally, you should not set `py_executable` to include shell logic like `uv python pin 3.9 && uv run`—Ray expects a single executable, not a shell command chain ([see docs](https://docs.ray.io/en/latest/ray-core/handling-dependencies.html#codecell15)). The recommended approach is to use `uv run` directly as the entrypoint and avoid setting `py_executable` unless you have a very specific advanced use case.

For best results, upgrade to Ray ≥2.47 and use the standard pattern:

```sh
ray job submit --working-dir . -- uv run -m myapp.mytask

```

This ensures both driver and workers use the uv environment as specified in your `pyproject.toml` ([docs](https://docs.ray.io/en/latest/ray-core/handling-dependencies.html#codecell15)). If you must use `py_executable`, only set it to `"uv run"` (not a shell chain), and ensure your cluster image has uv installed and the correct Python version.

Would you like a step-by-step breakdown or troubleshooting checklist?

Sources:

- [Ray issue: unsatisfiable uv dependencies lead to job ever hanging in pending](https://github.com/ray-project/ray/issues/56166)
- [Ray docs: Handling dependencies with uv](https://docs.ray.io/en/latest/ray-core/handling-dependencies.html#codecell15)
- [Ray docs: Use uv to manage Python dependencies](https://docs.anyscale.com/dependency-management/uv.md)

Hint: Mention @RunLLM in the post for followups.

---

<div class="post-metadata">

**Author:** ![bln\_prasad](https://sea2.discourse-cdn.com/flex020/user_avatar/discuss.ray.io/bln_prasad/32/8328_2.png) [@bln\_prasad](https://discuss.ray.io/u/bln_prasad)\
**Post date:** [November 24, 2025, 12:53pm UTC](https://discuss.ray.io/t/issues-with-setting-up-runtime-environment-using-uv-run/23328/3 "2025-11-24T12:53:30Z")

</div>

@RunLLM Thanks. I’ve switched to ray==2.52.0. with this i’m seeing below error on worker threads. not sure why only this import error(**(raylet) ModuleNotFoundError: No module named ‘ray’**) is coming. this is happening only when i include ““py\_executable”: “uv run”. Is this expected can you please suggest on this error?

0%| | 0/2 [00:00\<?, ?it/s]e[A(raylet, ip=10.6.121.149) Using CPython 3.10.19 interpreter at: /home/ray/anaconda3/bin/python3  
(raylet, ip=10.6.121.149) Creating virtual environment at: .venv  
(raylet, ip=10.6.121.149) warning: Failed to hardlink files; falling back to full copy. This may lead to degraded performance.  
(raylet, ip=10.6.121.149) If the cache and target directories are on different filesystems, hardlinking may not be supported.  
(raylet, ip=10.6.121.149) If this is intentional, set `export UV_LINK_MODE=copy` or use `--link-mode=copy` to suppress this warning.  
(raylet) There are tasks with infeasible resource requests that cannot be scheduled. See [Scheduling — Ray 2.52.0](https://docs.ray.io/en/latest/ray-core/scheduling/index.html#ray-scheduling-resources) for more details. Possible solutions: 1. Updating the ray cluster to include nodes with all required resources 2. To cause the tasks with infeasible requests to raise an error instead of hanging, set the ‘RAY\_enable\_infeasible\_task\_early\_exit=true’. This feature will be turned on by default in a future release of Ray.  
(autoscaler +40s) Removing 1 nodes of type workergroup (max number of worker nodes reached).  
(autoscaler +40s) Resized to 6 CPUs.  
(raylet) There are tasks with infeasible resource requests that cannot be scheduled. See [Scheduling — Ray 2.52.0](https://docs.ray.io/en/latest/ray-core/scheduling/index.html#ray-scheduling-resources) for more details. Possible solutions: 1. Updating the ray cluster to include nodes with all required resources 2. To cause the tasks with infeasible requests to raise an error instead of hanging, set the ‘RAY\_enable\_infeasible\_task\_early\_exit=true’. This feature will be turned on by default in a future release of Ray.  
(autoscaler +45s) Removing 1 nodes of type workergroup (max number of worker nodes reached).  
(autoscaler +45s) Resized to 6 CPUs.  
(raylet) [2025-11-24 04:40:34,327 E 473 473] (raylet) worker\_pool.cc:590: Some workers of the worker process(50830) have not registered within the timeout. The process is dead, probably it crashed during start.  
(raylet) There are tasks with infeasible resource requests that cannot be scheduled. See [Scheduling — Ray 2.52.0](https://docs.ray.io/en/latest/ray-core/scheduling/index.html#ray-scheduling-resources) for more details. Possible solutions: 1. Updating the ray cluster to include nodes with all required resources 2. To cause the tasks with infeasible requests to raise an error instead of hanging, set the ‘RAY\_enable\_infeasible\_task\_early\_exit=true’. This feature will be turned on by default in a future release of Ray.  
(autoscaler +50s) Removing 1 nodes of type workergroup (max number of worker nodes reached).  
(autoscaler +50s) Resized to 6 CPUs.  
(raylet) Traceback (most recent call last):  
(raylet) File “/home/ray/anaconda3/lib/python3.10/site-packages/ray/\_private/workers/default\_worker.py”, line 8, in  
(raylet) import ray  
(raylet) ModuleNotFoundError: No module named ‘ray’

---

<div class="post-metadata">

**Author:** ![RunLLM](https://sea2.discourse-cdn.com/flex020/user_avatar/discuss.ray.io/runllm/32/8249_2.png) [@RunLLM](https://discuss.ray.io/u/RunLLM)\
**Post date:** [November 24, 2025, 12:53pm UTC](https://discuss.ray.io/t/issues-with-setting-up-runtime-environment-using-uv-run/23328/4 "2025-11-24T12:53:58Z")

</div>

The error `(raylet) ModuleNotFoundError: No module named 'ray'` on worker startup means the Ray package is not installed in the Python environment that the worker process is using. This is a common issue when Ray creates a new virtual environment (e.g., `.venv`) for the runtime environment, but the Ray package itself is not present in that environment ([see discussion](https://github.com/ray-project/ray/issues/53848), [see also](https://discuss.ray.io/t/raylet-modulenotfounderror-no-module-named-ray-with-installed-ray/12625)).

To fix this, ensure that:

- The Ray package is included in your runtime environment dependencies (e.g., in your `pyproject.toml`, `requirements.txt`, or `runtime_env` pip/uv list).
- If using `uv`, make sure `ray` is listed as a dependency so it is installed in the worker’s environment.
- Avoid situations where Ray creates a new `.venv` that does not inherit the parent environment’s packages.

Example fix (for `runtime_env`):

```python
runtime_env = {
    "pip": ["ray==2.52.0", ...], # add other dependencies as needed
    ...
}

```

Or, if using `uv`, ensure `ray` is in your `pyproject.toml` dependencies.

Would you like more detailed troubleshooting steps?

Sources:

- [Ray issue: Ray 2.47 regression: All tasks hang when using uv](https://github.com/ray-project/ray/issues/53848)
- [Ray Discuss: raylet ModuleNotFoundError: No module named ‘ray’ with installed ray](https://discuss.ray.io/t/raylet-modulenotfounderror-no-module-named-ray-with-installed-ray/12625)

Hint: Mention @RunLLM in the post for followups.