Motivation

If you’ve set up an AWS Auto Scaling Group and need to deploy code to those running EC2 instances, this Capistrano task can streamline the process by targeting only healthy nodes in the group.

Requirements

  • aws-sdk-ec2 ~> 1
  • aws-sdk-autoscaling ~> 1
  • capistrano ~> 3

Installation

In your Rails app’s Gemfile, add:

gem 'capistrano3-asg-deploy'

Then run:

bundle install

Or install it yourself as:

gem install capistrano3-asg-deploy

Add this line to your application’s Capfile:

require 'capistrano/autoscaling_deploy'

Usage

Set credentials with AmazonEC2FullAccess permission in the capistrano deploy script / stage files add the following lines

set :aws_region, 'ap-northeast-1'
set :aws_access_key_id, 'YOUR AWS KEY ID'
set :aws_secret_access_key, 'YOUR AWS SECRET KEY'
set :aws_autoscaling_group_name, 'YOUR NAME OF AUTO SCALING GROUP NAME'
set :aws_deploy_roles, [:app, :web, :db]
set :aws_deploy_user, 'USER FOR SSH CONNECTION'
invoke 'autoscaling_deploy:setup_instances'

you can add more auto scaling configs to deploy to multiple auto scaling groups like a cluster

How it works:

This gem will fetch only running instances that have an auto scaling group name you specified

It will then reject the roles of :db and the :primary => true for all servers found but the first one

(from all auto scaling groups you have specified such as using more then once the auto scaling directive in your config — i.e cluster deploy)

this is to make sure a single working task does not run in parallel

you end up as if you defined the servers yourself like so:

server ip_address1, :app, :db, :web, :primary => true
server ip_address2, :app, :web
server ip_address3, :app, :web

Thanks, Happy Deployment