Katana ML Skipper Workflow

Returns queue name, based on task ID. This allows to route event to the correct queue, without hardcoding logic in the Web API.

File Structure

.
    api
    • models.py
    • router.py
    • workflow.json
    • workflow.py
  • endpoint.py
  • Dockerfile
  • README.md
  • workflow-pod.yaml
  • requirements.txt

Instructions

To run all services, check instructions in Getting Started

Run the service without Docker

  1. Install libraries
    2. pip install -r requirements.txt
  2. Start FastAPI
    3. uvicorn endpoint:app --port=5000 --reload
  3. Workflow FastAPI endpoints
    4. http://127.0.0.1:5000/docs

Build and run individual container

  1. Build container
    2. docker build --tag katanaml/skipper-workflow .
  2. Run container
    3. docker run -it -d --name skipper-workflow -p 5000:5000 katanaml/skipper-workflow:latest
  3. Workflow FastAPI endpoints
    4. http://127.0.0.1:5000/docs

Build and run Kubernetes Pod

  1. Create namespace
    2. kubectl create ns katana-skipper
  2. Create Pod
    3. kubectl apply -n katana-skipper -f workflow-pod.yaml
  3. Check Pod status
    4. kubectl get -n katana-skipper pods
  4. Describe Pod
    5. kubectl describe -n katana-skipper pods skipper-workflow
  5. Open Pod port for testing purposes
    6. kubectl port-forward -n katana-skipper deploy/skipper-workflow 5000:5000
  6. Open Pod logs
    7. kubectl logs -n katana-skipper -f -l app=skipper-workflow
  7. Test URL
    8. http://localhost:5000/docs
  8. Check Pod service
    9. kubectl get -n katana-skipper svc skipper-workflow
  9. Delete Deployment
    10. kubectl delete -n katana-skipper -f workflow-pod.yaml
  10. Delete all resources
    11. kubectl delete all --all -n katana-skipper

License

Licensed under the Apache License, Version 2.0. Copyright 2020-2021 Katana ML, Andrej Baranovskij. Copy of the license.