MySQL Connection Timeout Issues in Django ASGI Application

The "MySQL server has gone away" error typically occurs when a database connection has been idle for too long and the server closes it, but your application still tries to use it.

Causes of Your Issue

When using Django with ASGI, there's a key difference in how database connections are managed compared to WSGI:

1. ASGI's Long-Running Nature: Unlike WSGI where connections are typically closed after each request, ASGI applications can maintain long-running processes.

2. Connection Pooling: Your setting of CONN_MAX_AGE=600 isn't working because ASGI frameworks often maintain their own connection pools outside of Django's connection management.

Solutions

1. Implement Connection Health Checks

Add middleware or signals to verify connection health before queries:

PYTHONCopied
from django.db import connections
from django.db.utils import OperationalError

def check_connection():
    for conn in connections.all():
        try:
            c = conn.cursor()
            c.execute("SELECT 1")
            c.close()
        except OperationalError:
            conn.close()

2. Use a Database Connection Pool

Consider using a connection pooling library like django-db-connection-pool:

PYTHONCopied
DATABASES = {
    'default': {
        'ENGINE': 'dj_db_conn_pool.backends.mysql',
        'NAME': 'your_db_name',
        'USER': 'your_user',
        'PASSWORD': 'your_password',
        'HOST': 'your_host',
        'PORT': '3306',
        'POOL_OPTIONS': {
            'POOL_SIZE': 20,
            'MAX_OVERFLOW': 10,
            'RECYCLE': 300,  # Recycle connections after 5 minutes
        }
    }
}

3. Adjust MariaDB Configuration

You could also modify your MariaDB server settings:

wait_timeout = 31536000  # 1 year in seconds
interactive_timeout = 31536000

However, this is generally not recommended as it can lead to resource exhaustion.

4. Explicitly Close Connections in ASGI Lifecycle Events

In your ASGI application, explicitly close connections at appropriate lifecycle points:

PYTHONCopied
from django.db import connections

async def lifespan(scope, receive, send):
    message = await receive()
    if message["type"] == "lifespan.startup":
        # Application startup
        await send({"type": "lifespan.startup.complete"})
    elif message["type"] == "lifespan.shutdown":
        # Application shutdown - close all DB connections
        for conn in connections.all():
            conn.close()
        await send({"type": "lifespan.shutdown.complete"})

The most reliable approach is typically a combination of connection pooling with proper health checks to ensure stale connections are detected and refreshed.