Celery定时任务在DigitalOcean Ubuntu服务器未注册报错求助

Let's walk through why your send_email_task is throwing this KeyError('periodic_tasks.send_email_task',) error on your DO server, even though it works locally. Here are the most likely fixes to try:

1. Fix the Empty Broker Configuration (Critical!) #

First off, your Celery app is initialized with an empty broker string:

app = Celery('periodic_tasks', broker='')

Local environments might fall back to a default (like a local Redis/RabbitMQ instance you have running without thinking), but production servers won't. A missing broker prevents Celery from properly registering and routing tasks.

Fix:
Replace the empty string with a valid broker URL (e.g., Redis or RabbitMQ). For example, if using Redis:

app = Celery('periodic_tasks', broker='redis://localhost:6379/0')

Make sure the broker service is installed and running on your DO server too (e.g., sudo systemctl start redis-server).

2. Ensure Supervisord Runs the Worker from the Correct Directory #

If your periodic_tasks.py isn't in the working directory where supervisord starts the Celery process, the worker won't find the task module.

Fix:
Add the directory directive to your supervisord config for the Celery process, pointing to the folder containing periodic_tasks.py:

[program:celery-worker]
command=/path/to/your/venv/bin/celery -A periodic_tasks worker -B --loglevel=info
directory=/path/to/your/project/folder
user=your-ubuntu-user
autostart=true
autorestart=true
stdout_logfile=/var/log/celery/celery.log
stderr_logfile=/var/log/celery/celery.err

Also, note that using the full path to your virtual environment's Celery executable avoids conflicts with system-wide Celery installations.

3. Verify the Worker Actually Loads Your Task #

When the Celery worker starts, it logs a section like [tasks] that lists all registered tasks. Check your Celery logs (configured in supervisord above) — if periodic_tasks.send_email_task isn't listed there, the worker isn't loading the module.

Fixes:

• Add your project path to Python's sys.path if it's not already there. You can do this in periodic_tasks.py:

  import sys
  from pathlib import Path
  # Add the project root to Python's module search path
  sys.path.append(str(Path(__file__).parent.resolve()))
  

• Or prepend the PYTHONPATH to your supervisord command:

  command=PYTHONPATH=/path/to/your/project/folder /path/to/venv/bin/celery -A periodic_tasks worker -B --loglevel=info
  

4. Check for Case Sensitivity and Typos #

Linux filesystems are case-sensitive. Double-check that:

• Your file is named exactly periodic_tasks.py (not PeriodicTasks.py or periodic-tasks.py)
• The task name in beat_schedule ("periodic_tasks.send_email_task") matches the actual task's fully qualified name.

5. Test Separating Beat and Worker Processes #

Running Beat and Worker together with the -B flag can sometimes cause path or configuration conflicts. Try starting them separately to isolate the issue:

1. Stop the supervisord Celery process first.
2. Start the worker:

   cd /path/to/your/project && /path/to/venv/bin/celery -A periodic_tasks worker --loglevel=info
   

3. In a separate terminal, start Beat:

   cd /path/to/your/project && /path/to/venv/bin/celery -A periodic_tasks beat --loglevel=info
   

If this works, update your supervisord config to run two separate processes (one for worker, one for beat).

内容的提问来源于stack exchange，提问作者Konstantin Rusanov