Baike.dev
All toolsOnline toolsAI codingTrendingOpen sourceNewsSubmit
Log in
< Back to tools
A

async-rails

> 编程语言
Open source

async Rails 3 stack demo

462 stars0 likes2 views
WebsiteGitHub

About

async Rails 3 stack demo

Async Rails 3.1 stack demo

Important warning:

You should be aware when using fibers with Rails that you can get a stack overflow error if your stack grows
bigger than 4Kb (which is enough for most things though), this got even worse with the 3.1 release in which
you can easily overflow the stack, here is an example here.

Simple async demo stack with Rails 3.1 + EventMachine and Fibers.

  • Hit localhost:3000/widgets to do a 1s async mysql query
  • Hit localhost:3000/widgets/http to make an HTTP call back to /widgets - recursive! :-)
  • Hit localhost:3000/twitter to load a mounted async Sinatra app (reports latests rails 3 tweets)

Howto / example commits:

  • Modify your config.ru to include the Rack::FiberPool middleware: config.ru
  • Configure ActiveRecord to use async mysql driver: Gemfile, and database.yml
  • Use async HTTP fetching within Rails
  • Mount async Sinatra app

Requirements:

  • Ruby 1.9.x
  • Async app server (thin)
  • Rails 3.1

Environment setup:

  • rvm install 1.9.2
  • rvm gemset create async-rails
  • rvm use 1.9.2@async-rails
  • gem install rails thin

Starting up Rails:

  • bundle install
  • bundle exec thin -D start

Concurrency

ab -c 5 -n 10 http://127.0.0.1:3000/widgets/

    Concurrency Level:      5
    Time taken for tests:   2.740 seconds
    Complete requests:      10

We're running on a single reactor, so above is proof that we can execute HTTP+MySQL queries in non-blocking fashion on a single run loop / within single process:

  • AB opens 5 concurrent requests (10 total)
  • Each request to /widgets/http opens an async HTTP request to /widgets - aka, we ourselves spawn another 5 requests
  • Because the fiber pool is set to 10, it means we can process all 5 requests within ~1s (each mysql req takes 1s)
  • 10 requests finish in ~2s

So, keep in mind that the size of 'database pool' is basically your concurrency throttle. In example above, we spawn 10 requests, which open another 10 internally, so in total we process 20 req's in ~2s on a single thing server. Just as expected.

Benchmarks

Pushing the stack on my MBP (db pool = 250; fiber pool = 250; env = production; thin 1.2.7) results in:

    Concurrency Level:      220
    Time taken for tests:   10.698 seconds
    Complete requests:      2000
    Failed requests:        0
    Write errors:           0
    Total transferred:      470235 bytes
    HTML transferred:       12006 bytes
    Requests per second:    186.95 [#/sec] (mean)
    Time per request:       1176.777 [ms] (mean)
    Time per request:       5.349 [ms] (mean, across all concurrent requests)
    Transfer rate:          42.93 [Kbytes/sec] received

For full AB trace see this gist

Resources:

  • No callbacks, No threads - RailsConf 2010 Presentation
  • Rails performance needs an overhaul
  • Non-blocking ActiveRecord
  • Untangling evented code with Ruby fibers
  • Introducing Phat, and Async Rails app

Other:

  • Postgres + Async Rails driver
  • Fix: Asset Pipeline + Stack level too deep error

Issues· 4 open

View all issuesOpen on GitHub

No open issues yet, or sync has not completed.

> Tags

Ruby

No comments yet. Be the first to share.

> Details

PublishedAug 1, 2026
UpdatedSep 17, 2026
Category编程语言
PricingOpen source

> Related tools

T
TypeScript
JavaScript 的超集,为前端与全栈提供静态类型
P
Python
通用编程语言,广泛用于 Web、数据与 AI
G
Go
Google 推出的简洁高效系统语言