Flask应用迁移至Apache WSGI后无法启用多进程的问题求助

Hey there! Let's break down how to get your Flask app running with multiple processes under Apache mod_wsgi, and tackle those CPU overload issues you're facing.

The most likely reason you're stuck with a single process is using Apache's embedded mode for mod_wsgi, which ties your app directly to Apache's own worker processes and limits scaling control. Instead, use daemon mode—this lets you explicitly configure how many processes/threads your app uses, independent of Apache's core settings.

Update your VirtualHost configuration like this:

<VirtualHost *:80>
    ServerName your-domain.com

    # Define the daemon process pool
    WSGIDaemonProcess myapp 
        user=www-data group=www-data 
        processes=8  # Match your 8-core CPU for full hardware utilization
        threads=15   # Adjust based on your app's I/O vs CPU balance; 10-20 is typical
        python-path=/var/www/app-server:/path/to/your/virtualenv/lib/pythonX.X/site-packages

    # Tell Apache to use this daemon pool for your app
    WSGIProcessGroup myapp

    # Point to your WSGI file
    WSGIScriptAlias / /var/www/app-server/APPNAME.wsgi

    <Directory /var/www/app-server>
        Require all granted
    </Directory>

    # Optional: Serve static files directly from Apache (faster than Flask)
    Alias /static /var/www/app-server/static
    <Directory /var/www/app-server/static>
        Require all granted
    </Directory>
</VirtualHost>

• processes=8: Since you have an 8-core CPU, setting this to match your core count will let you fully leverage your hardware—critical for CPU-heavy workloads like your parsing app.
• python-path: Include your app directory and virtual environment's site-packages (if used) to avoid import errors.

If you prefer not to use daemon mode (not recommended for CPU-heavy Python apps), adjust Apache's Multi-Processing Module (MPM) settings to spawn more worker processes:

1. Check which MPM you're using:

   apache2ctl -M | grep mpm
   

2. If using mpm_prefork (common for embedded mode), edit /etc/apache2/mods-available/mpm_prefork.conf:

   StartServers       8
   MinSpareServers    8
   MaxSpareServers   16
   MaxRequestWorkers  128
   MaxConnectionsPerChild  0
   

   These settings ensure Apache starts with 8 processes (matching your cores) and scales up as needed.

Note: Daemon mode is still better for Python apps—it isolates your app from Apache's core processes and avoids GIL-related bottlenecks with threads.

Permissions often prevent mod_wsgi from spawning multiple processes or accessing app resources:

• Ensure your app directory and files are owned by the user/group specified in WSGIDaemonProcess (e.g., www-data:www-data):

  chown -R www-data:www-data /var/www/app-server
  chmod -R 755 /var/www/app-server
  chmod 644 /var/www/app-server/*.py /var/www/app-server/APPNAME.wsgi
  

• Confirm the www-data user has read/write access to any external resources your app uses (databases, files, APIs, etc.).

• Disable Debug Mode: Make sure app.debug = False in your code—Werkzeug's debug mode explicitly blocks multi-process setups for security reasons.
• Lazy Initialization: Each process initializes its own Flask app instance, so avoid heavy one-time setup (like loading large parsing models) in the global scope. Initialize these resources inside route functions or use lazy loading to avoid redundant work across processes.
• Avoid Shared State: Multi-process apps don't share memory, so don't rely on global variables for user data or cache. Use an external cache (like Redis) or database instead.

After updating configs, restart Apache:

sudo systemctl restart apache2

Then:

• Check for multiple mod_wsgi daemon processes:

  ps aux | grep mod_wsgi
  

  You should see 8 processes matching your processes setting.
• Monitor Apache logs for errors:

  tail -f /var/log/apache2/error.log
  

  Look for permission errors, import issues, or configuration typos.
• Test with traffic: Use tools like ab (Apache Bench) to simulate high load and confirm CPU usage is distributed across all cores, and your app stops stalling.

内容的提问来源于stack exchange，提问作者Sergey Kireyev