Capistrano Auto Scaling Group rolling deployments

Capistrano plugin for performing rolling updates to AWS Auto Scaling Groups using the instance refresh feature.

Instead of deploying to live servers, capistrano-asg-rolling will create a temporary instance for deployment and then trigger an instance refresh to perform a rolling update of the Auto Scaling Group(s). In more detail, during deployment it will:

• Launch an instance from the AMI defined in the Launch Template of the Auto Scaling Group(s).
• Deploy your application to the launched instances.
• After deployment, stop the instances and create an AMI for each instance.
• Create new Launch Template versions for the new AMIs.
• Trigger Instance Refresh on the Auto Scaling Group(s) to perform a rolling update.
• Delete any outdated Launch Template versions, AMIs and snapshots created by previous deployments.
• Terminate the no longer needed instances.

To better understand how Instance Refresh replaces instances, make sure to read the documentation: https://docs.aws.amazon.com/autoscaling/ec2/userguide/instance-refresh-overview.html

This gem depends on Auto Scaling Groups with Launch Templates. Using an Auto Scaling Group with a Launch Configuration is not supported, and will raise an Capistrano::ASG::Rolling::NoLaunchTemplate.

Instance refresh uses the desired configuration to update the Launch Template version of the Auto Scaling Group after a succesful deployment. Setting the Launch Template version to Latest on the Auto Scaling Group is not needed.

Please note that this gem works well for our configuration / use case, but it might not fit yours. Any feedback using GitHub issues / pull requests, is much appreciated.

Add this line to your application's Gemfile:

gem 'capistrano-asg-rolling'

And then execute:

$ bundle

Or install it yourself as:

$ gem install capistrano-asg-rolling

Add the plugin to Capistrano's Capfile:

# Capfile
require 'capistrano/asg/rolling'
install_plugin Capistrano::ASG::Rolling::Plugin

Below are the Capistrano configuration options:

Setup AWS credentials:

# config/deploy.rb
set :aws_access_key_id,     ENV['AWS_ACCESS_KEY_ID']
set :aws_secret_access_key, ENV['AWS_SECRET_ACCESS_KEY']
set :aws_region,            ENV['AWS_REGION']

After deployment, any outdated Launch Template versions, AMIs and snapshots will be deleted. By default, the number of keep_releases will be kept. Change this with:

# config/deploy.rb
set :asg_rolling_keep_versions, 10

Servers are added using their private IP address by default. Set to false to use the public IP address instead:

# config/deploy.rb
set :asg_rolling_use_private_ip_address, false

Verbose logging is enabled by default, set to false for less verbose logging:

# config/deploy.rb
set :asg_rolling_verbose, false

When launching an Instance, you can override any settings defined in the Launch Template with:

# config/deploy.rb
set :asg_rolling_instance_overrides, { instance_type: 'c5.large' }

You can make Capistrano wait until the instances in the autoscaling group have completed refreshing with:

# config/deploy.rb
set :asg_wait_for_instance_refresh, true
set :asg_instance_refresh_polling_interval, 30 # default

Enable or disable auto-rollback on instance refreshes (default: false):

# config/deploy.rb
set :asg_instance_refresh_auto_rollback, true

Specify the Auto Scaling Groups with the keyword autoscale instead of using the server keyword in Capistrano's stage configuration. Provide the name of the Auto Scaling Group and any properties you want to pass to the server:

# config/deploy/<stage>.rb
autoscale 'app-autoscale-group', user: 'deployer', roles: %w[db assets]
autoscale 'web-autoscale-group', user: 'deployer'

Now start a deployment with cap <stage> deploy and enjoy.

There might be cases where you just want to deploy your code to the servers in the Auto Scaling Group(s) without a rolling update.

You can configure rolling updates per autoscaling group by using the rolling option:

# config/deploy/<stage>.rb
autoscale 'app-autoscale-group', rolling: true    # default: use rolling deployment
autoscale 'web-autoscale-group', rolling: false   # override: use normal deployment

The instance refresh is triggered without specifying a value for the minimum / maximum healthy percentages. This means that either the default values will be used (minimum: 90%, maximum: 100%) or the percentages set in the instance maintenance policy for the Auto Scaling Group. You can tune both the minimum and maximum values to have more control about the desired capacity that must be healthy to proceed with replacing instances.

For example: reducing the minimum healthy percentage allows more instances to be terminated and new instances to be brought up at once during the instance refresh. eg. a value of 0 would terminate all instances in the autoscaling group and replace them at once.

You can configure the minimum healthy percentage per autoscaling group using the min_healthy_percentage option, and the maximum healthy percentage using the max_healthy_percentage option. Please note that if you specify max_healthy_percentage, you must also specify min_healthy_percentage.

# config/deploy/<stage>.rb
autoscale 'app-autoscale-group',                                                           # use default percentages or percentages set in the instance maintenance policy
autoscale 'web-autoscale-group', min_healthy_percentage: 75                                # allow 25% of instances to be terminated and replaced at once
autoscale 'web-autoscale-group', min_healthy_percentage: 100, max_healthy_percentage: 110  # allow for 10% above desired capacity during instance refresh

The rolling configuration of the stage has a side-effect: any Capistrano tasks you run, will also launch instances per Auto Scaling Group.

For example the command: cap production rails:console, will launch a new instance and run rails:console on that instance. While that can be useful, you often just want to run the task on the primary server. A solution is to create two stages with different rolling configurations, for example:

# config/deploy/production.rb
set :stage, :production

autoscale 'app-autoscale-group', rolling: false, user: 'deployer', roles: %w[db assets]
autoscale 'web-autoscale-group', rolling: false, user: 'deployer'

and:

# config/deploy/production_rolling.rb
set :stage, :production_rolling
set :rails_env, :production

autoscale 'app-autoscale-group', rolling: true, user: 'deployer', roles: %w[db assets]
autoscale 'web-autoscale-group', rolling: true, user: 'deployer'

With these two stages, you can run any tasks with cap production <task name> and rolling deployments with cap production_rolling deploy.

Do a test deployment: run the deploy task, but do not trigger the update ASG task and do not automatically terminate instances.

cap <stage> rolling:deploy_test

Just launch instance(s) defined in the Auto Scale Group(s) launch template. This instance is not attached to the Auto Scale Group and needs to be terminated manually.

cap <stage> rolling:launch_instances

Pick an instance in the Auto Scale Group(s), put it into standby and stop it. Then create an AMI and a new launch template. Then start the instance and put it into service again.

cap <stage> rolling:create_ami

You can filter any command to run on a specific Auto Scale Group by using the parameter asg_name.

For example, the command:

cap <stage> deploy asg_name=app-autoscale-group

Will do a deployment only on the Auto Scale Group with the name app-autoscale-group.

The following IAM permissions are required:

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Effect": "Allow",
            "Action": [
                "autoscaling:DescribeAutoScalingGroups",
                "autoscaling:DescribeAutoScalingInstances",
                "autoscaling:DescribeInstanceRefreshes",
                "autoscaling:EnterStandby",
                "autoscaling:ExitStandby",
                "autoscaling:StartInstanceRefresh",
                "ec2:CreateImage",
                "ec2:CreateLaunchTemplateVersion",
                "ec2:CreateTags",
                "ec2:DeleteLaunchTemplateVersions",
                "ec2:DeleteSnapshot",
                "ec2:DeregisterImage",
                "ec2:DescribeImages",
                "ec2:DescribeInstances",
                "ec2:DescribeLaunchTemplateVersions",
                "ec2:RunInstances",
                "ec2:StartInstances",
                "ec2:StopInstances",
                "ec2:TerminateInstances"
            ],
            "Resource": "*"
        }
    ]
}

After checking out the repo, run bin/setup to install dependencies. Then, run rake spec to run the tests. You can also run bin/console for an interactive prompt that will allow you to experiment.

To install this gem onto your local machine, run bundle exec rake install. To release a new version, update the version number in version.rb, and then run bundle exec rake release, which will create a git tag for the version, push git commits and tags, and push the .gem file to rubygems.org.

Bug reports and pull requests are welcome on GitHub at https://github.com/KentaaNL/capistrano-asg-rolling. This project is intended to be a safe, welcoming space for collaboration, and contributors are expected to adhere to the Contributor Covenant code of conduct.

The gem is available as open source under the terms of the MIT License.

Everyone interacting in the Capistrano::ASG::Rolling project’s codebases, issue trackers, chat rooms and mailing lists is expected to follow the code of conduct.