Skip to content

Is it considered to use another method to start the Ray cluster? #23

Description

@wjg13025152587

Hello! I found that currently, metaflow-ray is starting a Ray cluster through the Ray CLI in the pod started by jobset (Netflix/metaflow#365). This method currently does not support specifying the resources for Ray Node startup (such as nums_cpu, memory, nums_gpu, etc.). When using metaflow-ray on Kubernetes, this will cause the schedulable resources of Ray Cluster to be equal to the resources of the physical nodes. Currently, you can create a Ray Cluster in Kubernetes by using the Kuberay Operator. Does meraflow-ray also consider supporting the creation of Ray Cluster through the Kuberay Operator?

https://github.com/ray-project/kuberay

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Fields

    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions