5

I have a question related to this older question:

What exactly is a pre-fork web server model?

We had an internal discussion and a more senior dev insisted I was wrong. The disuccsion was around database connections.

When you deploy a wsgi application using gunicorn and you have 5 sync workers, I assumed that each worker is a completely separate process hosting the full wsgi application without any sharing with any other process or the master process.

In essence I wanted to explain that using a database connection pool probably doesn't make sense as each process will ever only have 1 active request and hence only ever needs 1 active connection. And that if you make a pool, all 5 workers will have their own completely separate connection pool so you end up with too many connections.

So my question is, are my assumption right or is the other dev right that claims the connection pool will be shared among synchronous workers?

beginner_
  • 7,230
  • 18
  • 70
  • 127
  • Your assumption is right. Whether Gunicorn is run with threads or processes, it will run your program's entrypoint on each worker. So if your entrypoint (for example your __init__.py or __main__.py) initializes a connection pool, each worker will be using its own connection pool. – WhyNotTryCalmer Jun 10 '23 at 08:06

0 Answers0