12-08-2022

  • managed to deploy ogbv-ml-rest to google cloud run
    • very simple gcloud run deploy --memory 8G --cpu 4
    • Query Command curl -X POST https://ogbv-ml-rest-■■■■■■-■■■■.a.run.app/predict -H 'Content-Type: application/json' -d '{"text":"■■■■ ■■■■ ■■■■■■■ ■■■■"}'
    • It should scale up and down automatically. It will also scale down to 0 when not being used which is a huge money saver
    • Since the requests to /predict go via uli-rest-api i might be able to keep this public for now and still migrate uli to gke
  • Conclusion for Uli
    • Lets try running the entire app on cloud run. No need to involve clusters and kubernetes
    • strategy
      • allow unauthenticated requests to uli-rest
      • uli-rest can communicate with ogbv-ml-rest via service-to-service authentication
      • uli-rest can communicate to a cloud-sql instance