CODEWITHSUNDEEP
pythongoogle-cloud-platformgcloudfastapigoogle-cloud-run
Khaled Ayman
Khaled Ayman

Reputation: 77

FastAPI hosted on Google cloud return 500 server error

I have a FastAPI API that takes an image, loads a Tensorflow model, and returns a prediction. Locally everything is working fine, but I want to host it on Google Cloud I followed a tutorial and did the following:

  1. created app.yaml file where the contents are
runtime: python39    
entrypoint: gunicorn -w 4 -k uvicorn.workers.UvicornWorker main:app
  1. requirements.txt having
fastapi
numpy
Pillow
pydantic
tensorflow
uvicorn
gunicorn
opencv-python

After that, I uploaded everything on Github, opened Google cloud, created a new project, activated cloud run and cloud build, cloned the git repo, and ran

gcloud app deploy app.yaml

I installed the requirements with

pip3 install -r requirements.txt

and finally launched the API using

gcloud app browse

What I get is either

upstream connect error or disconnect/reset before headers. reset reason: connection termination" or "Error: Server Error

or

The server encountered an error and could not complete your request.
Please try again in 30 seconds.

Upvotes: 1

Views: 1030

Answers (1)

Andres Fiesco Casasola
Andres Fiesco Casasola

Reputation: 801

As stated by Isabi, this was generated because the instance ran out of memory. Have you tried increasing the instance memory? Try changing the instance class in your app.yaml, 2, this will increase the memory limit according to the assigned class.

You can also use this stack answer to increase the instance memory

Upvotes: 1

Related Questions