Versions < 3 of ELBAS are no longer being maintained. I will only be maintaining the current feature-set which relies on Launch Templates and AWS SDK v3.
ELBAS was written to ease the deployment of Rails applications to AWS AutoScale groups. During your Capistrano deployment, ELBAS will:
- Deploy your code to each running instance connected to a given AutoScale group
- After deployment, create an AMI from one of the running instances
- Update the AutoScale group's launch template with the AMI ID
- Delete any outdated AMIs created by previous ELBAS deployments
Add to Gemfile, then bundle
:
gem 'elbas'
Add to Capfile:
require 'elbas/capistrano'
Setup AWS credentials:
set :aws_access_key, ENV['AWS_ACCESS_KEY_ID']
set :aws_secret_key, ENV['AWS_SECRET_ACCESS_KEY']
set :aws_region, ENV['AWS_REGION']
Instead of using Capistrano's server
method, use autoscale
instead in
deploy/<environment>.rb
(replace with your environment). Provide
the name of your AutoScale group instead of a hostname:
autoscale 'my-autoscale-group', user: 'apps', roles: [:app, :web, :db]
If you have multiple autoscaling groups to deploy to, specify each of them:
autoscale 'app-autoscale-group', user: 'apps', roles: [:app, :web]
autoscale 'worker-autoscale-group', user: 'apps', roles: [:worker]
Run cap production deploy
.
As of version 3, your AWS setup must use launch templates as opposed to launch
configurations. This allows ELBAS to simply create a new launch template version
with the new AMI ID after a deployment. It no longer needs to update your
AutoScale group or mess around with network settings, instance sizes, etc., as
that information is all contained within the launch template. Failure to use a
launch template will result in a Elbas::Errors::NoLaunchTemplate
error.
You can pass a block to autoscale
and return properties for any specific server.
The block accepts the server and the server's index as arguments.
For example, if you want to apply the :db
role to only the first server:
autoscale 'my-autoscale-group', roles: [:app, :web] do |server, i|
{ roles: [:app, :web, :db] } if i == 0
end
Returning nil
from this block will cause the server to use the properties
passed to autoscale
.
Returning anything but nil
will override the entire properties hash (as
opposed to merging the two hashes together).
You may need to SSH into your servers while debugging deployed code, and
not everyone has a jumpbox on a basic AutoScaling setup. ELBAS provides a command
that will list the ssh
command necessary to connect to each server in any given
environment:
cap production elbas:ssh
Output will be something like:
[ELBAS] Adding server: ec2-12-34-567-890.compute-1.amazonaws.com
[ELBAS] SSH commands:
[ELBAS] 1) ssh deploy@ec2-12-34-567-890.compute-1.amazonaws.com