Docker Swarm on AWS

This repo provisions a Docker Swarm Cluster on AWS using Packer and Terraform.

First export yout credentials and create the ami using packer. Provide the region variable when executing packer. The packer script uploads two files docker.conf and init.py. The first is for customize the docker service and the latter is a python script for initialize the cluster and add nodes to it.

export AWS_ACCESS_KEY_ID=your-access-key-here
export AWS_SECRET_ACCESS_KEY=your-secret-access-key-here
export AWS_DEFAULT_REGION=your-region

cd packer/
packer build -var 'region=us-east-1' docker.json

Write down the ami id generated.

Set the terraform variables like this:

export TF_VAR_manager_count=how-many-managers-do-you-want Ej: 1, 3, 5 (odd numbers recommended)
export TF_VAR_worker_count=how-many-workers-do-you-want Ej: 1,5,10,1000
export TF_VAR_key=your-key
export TF_VAR_ec2_ami=the-ami-key-generated-with-packer

Also check all the variables on the variables.tf file and modify as you wish. Now you can provision the infrastructure with Terraform

cd terraform/
terraform apply

A brief description of scripts is provided next:

• provider.tf defines the aws region.
• variables.tf contains all variables used on the scripts.
• securitygroups.tf defines the security groups for elastic load balancer and ec2 instances.
• iam.tf contains the iam role assigned to cluster nodes.
• loadbalancing.tf defines the elastic load balancer
• autoscaling.tf contains the launch configurations and autoscaling groups for worker and manager nodes.

Currently there are no outputs defined so you should login to AWS console to get nodes information.

TO-DO

• Think about how to initialize cluster and how worker nodes will automagically join to it
• Improve cluster initializer/join-nodes script
• Remove workers from cluster when replaced
• Deploy some CI/CD tools on cluster (Jenkins?)
• Add a Docker registry
• Add a proxy for accessing containers (Docker Flow Proxy?)
• Add a monitoring tool