How rails resolve multi-requests at the same time?

Question

Prepare for the test: sleep 10 in a action

Test: Open two tabs in the browser to visit the action

Result: When the second request is running, the first request finished and began rendering the view, but the view is still blank. After the second request finished too, the two requests finished rendering the view at the same time.

Conclusion: Rails is just one single instance. One request can only enter the action after the previous requests finish. But how to explain the response part? Why the multi-requests finish rendering the views at the same time?

Answer 1

WEBrick is multi-threaded but Rails developers hard-coded a mutex, so it can handle just one request at a time. You can monkey-patch Rails::Server and you are free to run a multi-threaded WEBrick.

Just note that WEBrick will be multithreaded only when config config.cache_classes = true and config.eager_load = true, which is typical to RAILS_ENV=production. This is because class reloading in development is not thread safe.

To get WEBrick fully multi-threaded in Rails 4.0, just add this to config/initializers/multithreaded_webrick.rb:

# Remove Rack::Lock so WEBrick can be fully multi-threaded. require 'rails/commands/server'  class Rails::Server   def middleware     middlewares = []     middlewares << [Rails::Rack::Debugger] if options[:debugger]     middlewares << [::Rack::ContentLength]      Hash.new middlewares   end end 

The offending code in rails/commands/server.rb that we got rid of is:

# FIXME: add Rack::Lock in the case people are using webrick. # This is to remain backwards compatible for those who are # running webrick in production. We should consider removing this # in development. if server.name == 'Rack::Handler::WEBrick'   middlewares << [::Rack::Lock] end 

It's not needed on Rails 4.2. It's concurrent out-of-the-box.

Answer 2

Are you using a WEBrick server? That must be because your server is a single threaded server and is capable of fulfilling one request at a time (because of the single worker thread). Now in case of multiple requests, it runs the action part of the request and before running the view renderer it checks to see if there are any pending requests. Now if 10 requests are lined up, it would first complete all of them before actually rendering the views. When all of these requests are completed, the views would now be rendered sequentially.

You can switch to Passenger or Unicorn server if you want multi-threaded environment.

Hope that makes sense.