What's the simplest way to design, build and test a concurrent but low throughput Flask app that uses flask-login?

[artgoldberg]

Hello flask-login folks

I'm building a Web API that will receive extremely little traffic. Occasionally one user will use it, rarely two users will use it concurrently, and three or more users will almost never use it concurrently.

The app uses flask-login, and stores users in an external database table. The user_loader and request_loader functions work:

from db import get_user # obtain a User object from the database
from flask import current_app

login_manager = flask_login.LoginManager()

@login_manager.user_loader
def user_loader(email):
  "obtain a user object from the user's ID (email) stored in the session"
  user = get_user(current_app, email)
  if user:
    return user
  return

@login_manager.request_loader
def request_loader(request):
  "obtain a user object from a request"
  email = request.form.get('email')
  user = get_user(current_app, email)
  if user:
    return user
  return

I've been using gunicorn, but the synchronous worker instances it launches for each request are not maintaining the status of users that are logged in.

This raises the questions:

• What's the simplest way to deploy this in production?
• What's a good way to unit test concurrent usage in the production configuration?
• Should the app check whether http clients support cookies, and raise an error if they don't?

Thanks
Arthur

[artgoldberg]

This problem was fixed by configuring gunicorn :

• Use 1 worker process
• Use gthread worker threads
• Create multiple threads (4) if needed