HOWTO deploy a docker container on aws without using elastic beanstalk or ec2 container service

Question

I'd like to use docker containers without having to use elastic beanstalk or ec2 container service. I'd like to upload a .zip file describing the container (like you do with elastic beanstalk) and have a generic ec2 instance run it using docker.

When looking into the user data section of a beanstalk-created ec2 instance running a docker container, I see a cloud-init script that downloads a big shell script that does all the setup ( Example ). I assume that everything that elastic beanstalk does can also be achieved manually by using ec2 instances and a user data script.

My question is: Could someone provide a minimal example for a user data script that

1. installs/configures docker
2. downloads the .zip file
3. runs my docker image

I'm familiar with auto scaling groups etc. and I'd like to get this setup running without using the beanstalk- or ec2 container service magic.

Answer 1

Basically, you need to install Docker and nginx (as web proxy) in your EC2 instance. And then, download the web-app archive and deploy it. This is what Elastic Beanstalk does.

For the basic/minimal user-data in order to deploy a single docker container web application:

#!/bin/bash

IMG_LABEL=myapp
APP_INIT_URL=https://s3.amazonaws.com/your-bucket-app/myapp-init.tar.gz

function prepare_instance {
  apt-get -y update
  apt-get -y install nginx
  curl -sSL https://get.docker.com/ | sh
  mkdir /opt
  curl -o /opt/deployer.sh http://169.254.169.254/latest/user-data
  chmod 775 /opt/deployer.sh
}

function download_app {
  curl -o /tmp/current.tar.gz $1
  rm -rf /opt/app
  mkdir -p /opt/app
  tar zxvf /tmp/current.tar.gz -C /opt/app
  rm /tmp/current.tar.gz
}

function build_image {
  docker tag ${IMG_LABEL}:latest ${IMG_LABEL}:prev || echo "No built app"
  docker build -t ${IMG_LABEL}:latest /opt/app
}

function run_container {
  APP_CID=$(docker run -d ${IMG_LABEL}:latest)
  APP_IP=$(docker inspect --format '{{ .NetworkSettings.IPAddress }}' ${APP_CID})
}

function setup_proxy {
  rm /etc/nginx/sites-enabled/*
  cat <<EOT > /etc/nginx/sites-enabled/app.conf
map \$http_upgrade \$connection_upgrade {
  default upgrade;
  ''      close;
}
upstream app.local {
  server ${APP_IP};
}
server {
  listen 80;
  location / {
    proxy_pass http://app.local;
    include /etc/nginx/proxy_params;
    proxy_http_version 1.1;
    proxy_set_header Upgrade \$http_upgrade;
    proxy_set_header Connection \$connection_upgrade;
  }
}
EOT
  service nginx reload
}

function destroy_previous {
  (docker ps -a --before="${APP_CID}" | awk '{ print $1,$2 }' | grep "${IMG_LABEL}" | awk '{print $1 }' | xargs -I {} docker stop {} | xargs -I {} docker rm {}) || echo "No previous container"
  docker rmi ${IMG_LABEL}:prev || echo "No previous image"
}

if [ ! -f /opt/deployer.sh ];
then
  prepare_instance
  download_app ${APP_INIT_URL}
else
  download_app $1
fi

build_image
run_container
setup_proxy
destroy_previous

In Elastic Beanstalk, there is an agent that listen to update request. But, to make it simple, we can call the above script to deploy a new web-app version via SSH:

ssh ubuntu@ec2-107-123-123-123.compute-1.amazonaws.com 'sudo /opt/deployer.sh https://s3.amazonaws.com/your-bucket-app/myapp-201510122341.tar.gz'

Note : I use EC2 instance with Ubuntu 14.04.

Answer 2

The user data is basically just a bash script that is run when the instance boots up first time.

If you want to look into setting up an instance from scratch on instance creation I suggest you have a look at CloudInit and how to use it in CloudFormation. http://docs.aws.amazon.com/AWSCloudFormation/latest/UserGuide/aws-resource-init.html

With CloudInit you can describe files to be placed, packages to be installed and services to be enabled for start at boot.