A GLI based command-line tool for deploying applications on AWS EC2 Container Service (ECS)
The wiki has all kinds of useful information on it so don't just rely on this README.
Amazon ECS presents a low barrier to entry for production-level docker applications. Combined with ECS's built-in blue-green deployment, Elastic Load Balancers, Autoscale Groups, and CloudWatch, one can theoretically set up a robust cluster that can scale to serve any number of applications in a short amount of time. The ECS GUI, CLI, and overall architecture are not the easiest to work with, however, so Broadside seeks to leverage the ECS ruby API to dramatically simplify and improve the configuration and deployment process for developers, offering a simple command line interface and configuration format that should meet most needs.
Broadside does not attempt to handle operational tasks like infrastructure setup and configuration, which are better suited to tools like terraform.
- Trigger ECS deployments
- Inject environment variables into ECS containers from local configuration files
- Launch a bash shell on container in the cluster
- SSH directly onto a host running a container
- Execute an arbitrary shell command on a container (or all containers)
- Tail logs of a running container
- Scale an existing deployment on the fly
Applications using broadside employ a configuration file that looks something like:
Broadside.configure do |config|
config.application = 'hello_world'
config.default_docker_image = 'lumoslabs/hello_world'
config.aws.ecs_default_cluster = 'production-cluster'
config.aws.region = 'us-east-1' # 'us-east-1 is the default
config.targets = {
production_web: {
scale: 7,
command: %w(bundle exec unicorn -c config/unicorn.conf.rb),
env_file: '.env.production'
predeploy_commands: [
%w(bundle exec rake db:migrate),
%w(bundle exec rake data:migrate)
]
},
# If you have multiple images or clusters, you can configure them per target
staging_web: {
scale: 1,
command: %w(bundle exec puma),
env_file: '.env.staging',
tag: 'latest', # Set a default tag for this target
cluster: 'staging-cluster', # Overrides config.aws.ecs_default_cluster
docker_image: 'lumoslabs/staging_hello_world' # Overrides config.default_docker_image
},
json_stream: {
scale: 1,
command: %w(java -cp *:. path.to.MyClass),
# This target has a task_definition and service config which you use to bootstrap a new AWS Service
service_config: { deployment_configuration: { minimum_healthy_percent: 0.5 } },
task_definition_config: { container_definitions: [ { cpu: 1, memory: 2000, } ] }
}
}
end
From here, developers can use broadside's command-line interface to initiate a basic deployment and launch the
configured command
as an ECS Service:
bundle exec broadside deploy full --target production_web --tag v.1.1.example.tag
In the case of an error or timeout during a deploy, broadside will automatically rollback to the latest stable version. You can perform manual rollbacks as well through the command-line.
First, install broadside by adding it to your application Gemfile
:
gem 'broadside'
Then run
bundle install
You can now run the executable in your app directory:
bundle exec broadside --help
You may also generate binstubs using
bundle binstubs broadside
Alternatively, you can install broadside using:
gem install broadside
For full application setup, see the detailed instructions in the wiki.
Use the --debug
switch to enable stacktraces and debug output.
Pull requests, bug reports, and feature suggestions are welcome!
Before starting on a contribution, we recommend opening an issue or replying to an existing one to give others some initial context on the work needing to be done.
Specs must pass on pull requests for them to be considered.
Broadside has a lot of tests for most of its behaviors - just run
bundle exec rspec
in the broadside directory. Don't open pull requests without passing specs.