Skip to content

Allow users to debug Cortex containers #2077

@vishalbollu

Description

@vishalbollu

Currently Cortex containers are deeply integrated with the Cortex operator and Cortex cluster. They make specific assumptions about how the project code and models are mounted, environment variables, ports. In addition the api spec consumed by the Cortex containers likely requires the need for the Cortex operator. The aforementioned reasons make it extremely difficult to run Cortex containers locally, making it difficult to test for users building their containers.

Add a guide/command to allow users to run Cortex containers locally.

Considerations:

  • Add defaults to environment variables
  • Only initialize statsd client when the necessary (ie in the cluster)
  • How do we deal with multi container APIs such TensorFlow serving? (run the containers separately?)
  • How do we deal with TaskAPI/BatchAPI/AsyncAPIs? (deploy multiple containers, a realtime api container to process the request followed by the worker container?)

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    No projects

    Milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions