1572147 : Usage of maxwell jupyterhub service

Created: 2026-01-29T09:57:36Z - current status: new

Here is the anonymized and summarized version of the reported issue:

Summary of the Issue

A user executed a command (uvicorn) on a shared compute node (max-wne006) of the Maxwell cluster, which resulted in the node crashing. The process consumed excessive resources, disrupting other active user sessions on the same node.

The command in question was:

uvicorn [APPLICATION]:app --host 0.0.0.0 --port [PORT]

This was run from a virtual environment located in the user’s project directory.

Suggested Solution/Next Steps

  1. Resource Management:
  2. Avoid running resource-intensive services (e.g., web servers like uvicorn) directly on shared compute nodes. These nodes are intended for batch jobs or interactive sessions with controlled resource usage.

  3. If such services are required, consider:

    • Using dedicated partitions (e.g., maxgpu or maxcpu) with proper resource requests (CPU, memory, GPU).
    • Running the service in a containerized environment (e.g., Singularity) with resource limits.

  4. Alternatives for Web Services:

  5. For development/testing, use JupyterHub (max-jhub.desy.de) or local environments.

  6. For production-grade services, request a dedicated node or explore DESY’s cloud/container services (e.g., UCloud).

  7. Communication:

  8. If unsure about resource requirements, contact the Maxwell support team via the ticket system: maxwell.service@desy.de.

  9. Subscribe to the maxwell-user mailing list (link) for updates and best practices.

  10. Acknowledgment:

  11. Ensure proper acknowledgment of Maxwell resources in publications (see documentation).

Sources Used

  1. Maxwell Documentation: Contact & Mailing List
  2. JupyterHub on Maxwell
  3. Maxwell Services List
  4. Access to Maxwell Resources