多VM负载均衡下Django+Celery+RabbitMQ集群部署可行性及配置指导

安全阿岳

2026-5-20

Absolutely, this deployment setup is totally feasible—and it’s actually a standard approach for scaling Django + Celery workloads while keeping things reliable. Let’s start with why this works, then dive into the step-by-step config you need to pull it off.

Is This Deployment Feasible?

Yes, 100%. Here’s the breakdown:

RabbitMQ clusters deliver high availability and load balancing for your message queue. If one node goes down, others take over, so your Celery workers never lose access to tasks.
Running one Celery worker per Django VM is a clean way to distribute workload across your existing infrastructure. Each worker can handle tasks generated by its local Django instance (or any instance, since they all connect to the same RabbitMQ cluster).
Celery’s default task delivery mechanism ensures a single task is only picked up by one worker—RabbitMQ won’t dispatch the same task to multiple consumers as long as your cluster and Celery config are set up correctly. The only edge case to guard against is accidental duplicate task submissions (more on that later).

Step-by-Step Configuration Guide

1. Set Up the RabbitMQ Cluster

Let’s assume you’re using 3 VMs for the RabbitMQ cluster (a minimum of 2 is recommended for high availability).

Prerequisites

Install RabbitMQ on all 3 VMs (use your OS’s package manager, e.g., apt install rabbitmq-server on Debian/Ubuntu).
Ensure all nodes can communicate over ports 5672 (AMQP), 4369 (Erlang port mapper), and 25672 (RabbitMQ inter-node communication).

Configure Erlang Cookie (Critical!)

RabbitMQ uses an Erlang cookie to authenticate cluster nodes—it must be identical on all nodes:

On the first node (rabbit-01), locate the cookie file at /var/lib/rabbitmq/.erlang.cookie (it’s a hidden file).

Copy this file to rabbit-02 and rabbit-03, then fix permissions:

scp /var/lib/rabbitmq/.erlang.cookie rabbit-02:/var/lib/rabbitmq/
chmod 600 /var/lib/rabbitmq/.erlang.cookie
chown rabbitmq:rabbitmq /var/lib/rabbitmq/.erlang.cookie

Form the Cluster

Stop RabbitMQ on rabbit-02 and rabbit-03:
```
sudo systemctl stop rabbitmq-server
```

Reset nodes to clear existing state:

sudo rabbitmqctl stop_app
sudo rabbitmqctl reset

Join them to rabbit-01:

sudo rabbitmqctl join_cluster rabbit@rabbit-01
sudo rabbitmqctl start_app

Verify cluster status from any node:
```
sudo rabbitmqctl cluster_status
```

Configure Mirror Queues (High Availability)

To avoid task loss if a RabbitMQ node fails, mirror all Celery queues across the cluster:

sudo rabbitmqctl set_policy ha-all "^celery" '{"ha-mode":"all", "ha-sync-mode":"automatic"}'

This policy applies to all queues starting with celery (Celery’s default queue name) and mirrors them to every cluster node.

Create a Dedicated Celery User

For security, avoid using the default guest user:

sudo rabbitmqctl add_user celery your_secure_password
sudo rabbitmqctl set_user_tags celery management
sudo rabbitmqctl set_permissions -p / celery ".*" ".*" ".*"

2. Configure Django + Celery

Install Dependencies

In your Django project’s virtual environment:

pip install celery

Project Initialization

Assume your Django project is named myproject. Create a celery.py file in the project root (next to settings.py):

import os
from celery import Celery

# Set Django settings as default for Celery
os.environ.setdefault('DJANGO_SETTINGS_MODULE', 'myproject.settings')

app = Celery('myproject')

# Load config from Django settings (prefix with CELERY_)
app.config_from_object('django.conf:settings', namespace='CELERY')

# Auto-discover tasks from all Django apps
app.autodiscover_tasks()

Update myproject/__init__.py to load Celery with Django:

from .celery import app as celery_app

__all__ = ('celery_app',)

Update Django Settings

Add these configurations to settings.py:

# Connect to RabbitMQ cluster
CELERY_BROKER_URL = 'amqp://celery:your_secure_password@rabbit-01:5672,rabbit-02:5672,rabbit-03:5672//'

# Optional: Track task results (use Django DB or Redis)
CELERY_RESULT_BACKEND = 'django-db'

# Critical settings to prevent duplicate task execution
CELERY_TASK_ACKS_LATE = True
CELERY_WORKER_PREFETCH_MULTIPLIER = 1

CELERY_TASK_ACKS_LATE: Workers only acknowledge tasks after completion (instead of when received). If a worker crashes mid-task, RabbitMQ requeues it for another worker.
CELERY_WORKER_PREFETCH_MULTIPLIER: Limits workers to prefetch 1 task at a time, ensuring fair distribution and preventing task hoarding.

Start Celery Workers on Each Django VM

Run this command on every Django VM (use systemd for production to keep workers running):

celery -A myproject worker --loglevel=info

Create a systemd service file (/etc/systemd/system/celery.service) for auto-start:

[Unit]
Description=Celery Service
After=network.target

[Service]
User=your_django_user
Group=your_django_group
WorkingDirectory=/path/to/your/django/project
Environment="PATH=/path/to/your/virtualenv/bin"
ExecStart=/path/to/your/virtualenv/bin/celery -A myproject worker --loglevel=info

Restart=always

[Install]
WantedBy=multi-user.target

Enable and start the service:

sudo systemctl daemon-reload
sudo systemctl enable celery.service
sudo systemctl start celery.service

3. Prevent Accidental Duplicate Task Submissions

Even with perfect queue config, duplicate tasks can happen if your Django app gets duplicate requests (e.g., load balancer retries). Fix this with:

Idempotent tasks: Design tasks to produce the same result even if run multiple times. For example, use a unique identifier (user ID + request ID) to check if the task was already executed before processing.
Explicit task IDs: When submitting tasks, generate a unique ID to avoid duplicates:
```
from myapp.tasks import my_task
import uuid

task_id = f"my_task_{user_id}_{uuid.uuid4()}"
my_task.apply_async(args=[user_id], task_id=task_id)
```
Celery rejects duplicate task IDs, so only one instance of the task will be processed.

4. Monitoring & Maintenance

RabbitMQ Management Console: Enable the plugin to monitor cluster health:
```
sudo rabbitmq-plugins enable rabbitmq_management
```
Access it at http://rabbit-01:15672 using your celery user credentials.

Celery Flower: Install Flower to track worker status and task execution:

pip install flower

Run it with:

celery -A myproject flower --broker=amqp://celery:your_secure_password@rabbit-01:5672,rabbit-02:5672,rabbit-03:5672//

Access the dashboard at http://flower-host:5555.

内容的提问来源于stack exchange，提问作者shubham003