适用于Rails的Phusion Passenger应用预加载器

Question

I recently switched to using Phusion Passenger 4.0.50 in replacement for Unicorn. The reason is because Passenger integrates well with Nginx and can also power Node.js app. However, what I am wondering is if I should do anything about the external connections (with Postgresql, Redis, Memcached) like I did with Unicorn.

I found some code on the Passenger git such as this one . But this code seems to belong to Passenger rather than a Rails app.

As of now I am using Ruby 2.1.2, and Rails 4.1.6. Would such work to handle external connections still required? If it is, how should I do it?

Thanks.

Update:

Per this thread , it seems that normal ActiveRecord is taken cared of automatically by Passenger.

Answer 1

Phusion Passenger author here. Yes you must do something about external connections. The "smart spawning" concept in Phusion Passenger is exactly the same as the "preload_app on" concept in Unicorn. We have an entire section in the documentation explaining how it works and what the caveats are (specifically, about external connections): https://www.phusionpassenger.com/documentation/Users%20guide%20Nginx.html#spawning_methods_explained

The only exception is the Rails default ActiveRecord connection. We automatically reestablish that, because it counts for over 90% of the use cases.

Answer 2

Passenger starts multiple sepearate worker processes, each of them loading the rails application. So each webrequest is handled by a completely isolated process within one thread. This means you don't have to care about connection pools at all. For example, in case of ActiveRecord, there are 5 concurrent connections possible by default (can be adjusted) and in your case you allways have only one connection (per process)