[Celery + RabbitMQ] Worker cannot connect to broker

1

I have a RabbitMQ web service and a Celery background worker.

RabbitMQ is running as expected but the Celery worker cannot connect and I receive this error message:

consumer: Cannot connect to amqp://<USER>:**@<HOST>:<PORT>//: timed out.
Trying again in 2.00 seconds... (1/100)

Both services are on the same region and account (they are defined in the same blueprint).

I am creating the Celery broker URL like this:

broker = f"amqp://{username}:{password}@{hostport}//"

username and password are set using an env group and appear to be correct.

hostport is set in the blueprint like this:

      - key: BROKER_HOSTPORT
        fromService:
          type: web
          name: rabbitmq
          property: hostport

Full render.yaml:

services:
  - type: web
    name: rabbitmq
    region: ohio
    env: docker
    rootDir: rabbitmq
    envVars:
      - key: RABBITMQ_ERLANG_COOKIE
        generateValue: true
      - fromGroup: broker-settings
    disk:
      name: rabbitmq
      mountPath: /var/lib/rabbitmq
      sizeGB: 10
  - type: worker
    name: ai-worker
    region: ohio
    env: python
    buildCommand: "pip install -r worker/requirements.txt"
    startCommand: "make worker"
    envVars:
      - key: BROKER_HOSTPORT
        fromService:
          type: web
          name: rabbitmq
          property: hostport
      - key: CELERY_RESULT_BACKEND
        fromService:
          type: redis
          name: celery-redis
          property: connectionString
      - key: PYTHON_VERSION
        value: 3.9.13
      - key: OPENAI_API_KEY
        sync: false
      - key: PROMPTLAYER_API_KEY
        sync: false
      - key: MEDIUM_INTEGRATION_TOKEN
        sync: false
      - fromGroup: broker-settings
  - type: web
    name: flower
    region: ohio
    plan: free
    env: python
    buildCommand: "pip install -r worker/requirements.txt"
    startCommand: "make flower"
    envVars:
      - key: BROKER_HOSTPORT
        fromService:
          type: web
          name: rabbitmq
          property: hostport
      - key: PYTHON_VERSION
        value: 3.9.13
      - fromGroup: broker-settings
  - type: redis
    name: celery-redis
    region: ohio
    plan: starter # we choose a plan with persistence to ensure tasks are not lost upon restart
    ipAllowList: [] # only allow internal connections

Celery worker client.py:

from celery import Celery, group
from dotenv import load_dotenv
import os
load_dotenv()

broker = os.environ["CELERY_BROKER_URL"] if "CELERY_BROKER_URL" in os.environ else None

if not broker:
    hostport = os.environ["BROKER_HOSTPORT"] if "BROKER_HOSTPORT" in os.environ else "localhost:5672"
    username = os.environ["RABBITMQ_DEFAULT_USER"] if "RABBITMQ_DEFAULT_USER" in os.environ else "guest"
    password = os.environ["RABBITMQ_DEFAULT_PASS"] if "RABBITMQ_DEFAULT_PASS" in os.environ else "**"
    broker = f"amqp://{username}:{password}@{hostport}//"

backend = os.environ["CELERY_RESULT_BACKEND"] if "CELERY_RESULT_BACKEND" in os.environ else "redis://"

client = Celery('client',
                broker=broker,
                backend=backend,
                include=['worker.tasks.generate', 'worker.tasks.publish'],
                setup_defaults={"time_limit": 120, "soft_time_limit": 120})


# Optional configuration, see the application user guide.
client.conf.update(
    result_expires=3600,
)

if __name__ == '__main__':
    client.start()

Any suggestions?

2

Ended up switching the broker to a Redis service and it works but still curious why RabbitMQ did not work.

3

Hi,

You didn’t go into too much detail on your RabbitMQ setup, but if it’s similar to Render’s example repo it’s deployed to a Web Service and multiple processes bind to different ports. I suspect hostport is providing the HTTP admin endpoint, where you’ll want the TCP/AMQP port. Maybe try using just the host in your env var and set the Rabbit AMQP port manually to its own env var, or direct in the code, e.g.:

broker = f"amqp://{username}:{password}@{host}:5672//"

Hope that helps

Alan

4

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.