Description
What happened + What you expected to happen
When I ray up a cluster which installs run ray via uv instead of pip install, it breaks the dashboard, dashboard exists with segfault.
Versions / Dependencies
Ray: 2.46.0
OS: ubuntu
Python: 3.10.12
Reproduction script
in ray-cluster.yaml
setup_commands:
- sudo apt-get update -y
- sudo snap install astral-uv --classic
- uv init .
- uv add ray[default,serve]
head_start_ray_commands:
- uv run ray stop
- uv run ray start --head --autoscaling-config=~/ray_bootstrap_config.yaml --dashboard-host=0.0.0.0
- uv run serve start --http-host 0.0.0.0
result, Ray + Serve start correctly, except that the dashboard is unreachable. Upon closer inpsection, it seems it seg faults:
additional information, from logs:
_(dashboard exits with segfault)
[2025-06-06 01:14:12,039 E 3616 3664] (raylet) agent_manager.cc:87: The raylet exited immediately because one Ray agent failed, a>
The raylet fate shares with the agent. This can happen because
The version of grpcio doesn’t follow Ray’s requirement. Agent can segfault with the incorrect grpcio version. Check the grp>
The agent failed to start because of unexpected error or port conflict. Read the log `cat /tmp/ray/session_latest/logs/{dashboa>
The agent is killed by the OS (e.g., out of memory)._
Issue Severity
Medium: It is a significant difficulty but I can work around it.