Flask Socket-IO Emitting Across Google Cloud Run Container Instances

[charbelmarche33]

I have been developing a web application that uses a flask web server, flask socket-io, and flask sessions. Previously I was deploying via a docker image to Google Cloud Run. Google Cloud Run increases the number of active containers based on traffic, and to allow for session permeance, I am newly deploying to the web with gunicorn, redis, eventlet.

Working with redis and eventlet has raise a problem with emitting with flask socket-io however. The following message is logged when trying to emit:

Cannot publish to redis... giving up

This causes emitting to not work.

Upon searching the web, I believe my problem is because Redis is not set up properly to allow the Google Cloud Run containers to emit to with one another. I've been playing with the configuration for days, but cannot seem to figure it out. Does anyone know why this issue is arising? I tried to follow the docs as closely as possible.

Here is my docker file

FROM python:3.8.12
WORKDIR .
RUN apt-get update && apt-get install -y \
  redis-server 
COPY requirements.txt /tmp/requirements.txt
COPY . .
RUN pip install --upgrade pip
RUN pip install --no-cache-dir -r /tmp/requirements.txt
ENV PORT 8080
#EXPOSE $PORT
CMD exec gunicorn --worker-class eventlet --bind :$PORT --workers 1 --threads 8 --timeout 0 server:app

The set up within my server.py file

import eventlet
eventlet.monkey_patch()
# other imports
# ...
# Initiate session
Session(app)

# Initiate socket IO object
socketio = SocketIO(app, message_queue='redis://')
# I additionally tried socketio = SocketIO(app, message_queue='redis://0.0.0.0:6379')

# Example of how I emit within a path
@app.route('/path', methods=['GET'])
def functionThatEmits():
    # some stuff
    socketio.emit(someval, someval, namespace='/anamespace', broadcast=True)
    # more stuff

# How the app is run
if __name__ == "__main__":
    socketio.run(app, host='0.0.0.0', port=8080, debug=False) # For running on gcloud

The versions of redis, eventlet, flask-socketio, flask, and gunicorn are below.

eventlet==0.33.3
Flask==2.3.2
flask_session==0.5.0
Flask_SocketIO==5.1.1
redis==5.0.0 
gunicorn==21.2.0

Thank you all for your time.

[miguelgrinberg]

I don't think you can mix threads and eventlet in the same worker. Try without the threads option.

[charbelmarche33]

Oh wow thank you so much for the help, really appreciate it.

I ended up trying redis memorystore from google cloud and it seems to have solved my problem. For anyone who comes across this in the future follow along here and use the following for the gunicorn command.

gunicorn --worker-class eventlet --bind :$PORT --workers 1 --timeout 0 server:app

Define socket IO like this:

socketio = SocketIO(app, message_queue=f'redis://{redis_host}:{redis_port}')

Lastly run the app like so:

socketio.run(app, host='127.0.0.1', port=8080, debug=False)

Leave the following the eventlet monkeypatch at the top of the server.py. Make sure you import redis in your server.py. See below:

import eventlet
eventlet.monkey_patch()

import redis

[miguelgrinberg]

Your solution is going to let you have up to 7 concurrent clients. For most people this is going to be insufficient. Eventlet is what helps the application scale to much larger levels.

[charbelmarche33]

Updated the original comment!