MongoDB: What is connection pooling and timeout?

Question

So my Passenger spins up 5 instances of my Rails app

I connect to MongoDB using Connection.new("localhost", 3000, :pool_size => 1, :timeout => 5)

Why would I need a "pool of connections" if I only incur overhead when starting up my Rails app, not per request? Why would a single process need more than 1 connection?

And what is the purpose of the timeout? Why would I want the connection to timeout? Shouldn't it be persist through the Rails process' lifetime?

So confused...

This question is specific to Ruby and Mongo but I guess it applies to other languages/databases.

Answer 1

You don't need to use connection pooling here. As you're using Passenger, just make sure that each instance uses a separate connection by catching the start_worker_process event. This is documented in the driver README.

Connection pooling can be useful for certain multi-threaded apps. pool_size is the max number of concurrent threads the connection will handle, and timeout is the maximum number of seconds a thread can wait for an available socket before an exception is thrown.