I have setup gunicorn with 3 workers 30 worker connections and using eventlet worker class. It is setup behind Nginx. After every few requests, I see this in the logs.
For me, the solution was to add --timeout 90 to my entrypoint, but it wasn't working because I had TWO entrypoints defined, one in app.yaml, and another in my Dockerfile. I deleted the unused entrypoint and added --timeout 90 in the other.