Reflex + RabbitMQ RPC Timeout in Docker Setup

[DmolinaElectroingenium]

Hi everyone, I'm running into a specific issue using Reflex inside Docker containers when making long-running RPC calls over RabbitMQ. I’m sharing all details in case anyone has a more robust solution or insight.

🧱 Setup Overview:

• RPC Client (requester): runs in a Docker container with Reflex (reflex run --backend-only)
• RPC Server (responder): runs in a separate container, also using Python + pika
• Communication handled via RabbitMQ on a shared Docker network

📌 The Problem:
When an RPC call takes longer than ~90 seconds, I get this error from the RPC server:

• ConnectionResetError: [Errno 104] Connection reset by peer

This happens when the server tries to send the response back to the client.
At the same time, the Reflex backend logs this:

• [CRITICAL] WORKER TIMEOUT (pid:XX)

Which means: Gunicorn is killing the Reflex backend worker before the RPC response arrives, so the socket is gone, and the server can’t deliver the result.

🧪 What I’ve tried:

• Change rxconfig.py by adding the timeout parameter like:
  import reflex as rx

config = rx.Config(
app_name="Interfaz", # The name of the app (should match the name of the app directory).
api_url="http://localhost:8000", #"http://localhost:8000", The backend url the frontend will connect to. This must be updated if the backend is hosted elsewhere, or in production.
# frontend_path=reflex_frontend, # The path the frontend will be hosted on. This must be updated if the frontend is hosted elsewhere, or in production.
# deploy_url=reflex_frontend, # The url the frontend will be hosted on. This must be updated if the frontend is hosted elsewhere, or in production.
frontend_port=3000, # The port the frontend will be hosted on.
backend_port=8000, # The port the backend will be hosted on.
# cors_allowed_origins=[reflex_frontend], # List of origins that are allowed to connect to the backend API.
timeout= 300, # TEST
gunicorn_workers= 1, # The number of gunicorn workers to use. This should be set to the number of CPU cores available.
gunicorn_worker_class = "uvicorn.workers.UvicornWorker"
)
but is not working.

I’d like to:

• Allow long-running RPC calls without Gunicorn killing the backend
  Has anyone successfully increased the Gunicorn timeout ?
  Any suggestions or patterns that work better for long-lived RPC interactions with Reflex?

Thanks in advance! 🙏