Docker for AWS Setup & PrerequisitesEstimated reading time: 7 minutes
If your account has the proper permissions, you can use the blue button from the stable or beta channel to bootstrap Docker for AWS using CloudFormation. For more about stable and beta channels, see the FAQs.
|Stable channel||Beta channel|
|This deployment is fully baked and tested, and comes with the latest GA version of Docker Engine.
This is the best channel to use if you want a reliable platform to work with.
These releases follow a version schedule with a longer lead time than the betas, synched with Docker Engine releases and hotfixes.
|This deployment offers cutting edge features and comes with the experimental version of Docker Engine, described in the Docker Experimental Features README on GitHub.
This is the best channel to use if you want to experiment with features under development, and can weather some instability and bugs. This channel is a continuation of the beta program, where you can provide feedback as the apps evolve. Releases are typically more frequent than for stable, often one or more per month.
We collect usage data on betas across the board.
|Deploy Docker for AWS (stable)||Deploy Docker for AWS (beta)|
Deploy Docker for AWS (beta)
uses your existing VPC
There are two ways to deploy Docker for AWS:
- With a pre-existing VPC
- With a new VPC created by Docker
We recommend allowing Docker for AWS to create the VPC since it allows Docker to optimize the environment. Installing in an existing VPC requires more work.
Create a new VPC
This approach creates a new VPC, subnets, gateways and everything else needed in order to run Docker for AWS. It is the easiest way to get started, and requires the least amount of work.
All you need to do is run the CloudFormation template, answer some questions, and you are good to go.
Install with an Existing VPC
If you need to install Docker for AWS with an existing VPC, you need to do a few preliminary steps. See recommended VPC and Subnet setup for more details.
Pick a VPC in a region you want to use.
Make sure the selected VPC is setup with an Internet Gateway, Subnets, and Route Tables
You need to have three different subnets, ideally each in their own availability zone. If you are running in a region with only two Availability Zones, you will need to add more than one subnet into one of the availability zones. For production deployments we recommend only deploying to regions that have three or more Availability Zones.
When you launch the docker for AWS CloudFormation stack, make sure you use the one for existing VPCs. This template will prompt you for the VPC and subnets that you want to use for Docker for AWS.
- Access to an AWS account with permissions to use CloudFormation and creating the following objects. Full set of required permissions.
- EC2 instances + Auto Scaling groups
- IAM profiles
- DynamoDB Tables
- SQS Queue
- VPC + subnets and security groups
- CloudWatch Log Group
- SSH key in AWS in the region where you want to deploy (required to access the completed Docker install)
- AWS account that support EC2-VPC (See the FAQ for details about EC2-Classic)
For more information about adding an SSH key pair to your account, please refer to the Amazon EC2 Key Pairs docs
Docker for AWS is installed with a CloudFormation template that configures Docker in swarm-mode, running on instances backed custom AMIs. There are two ways you can deploy Docker for AWS. You can use the AWS Management Console (browser based), or use the AWS CLI. Both have the following configuration options.
Pick the SSH key that will be used when you SSH into the manager nodes.
The EC2 instance type for your worker nodes.
The EC2 instance type for your manager nodes. The larger your swarm, the larger the instance size you should use.
The number of workers you want in your swarm (0-1000).
The number of Managers in your swarm. You can pick either 1, 3 or 5 managers. We only recommend 1 manager for testing and dev setups. There are no failover guarantees with 1 manager — if the single manager fails the swarm will go down as well. Additionally, upgrading single-manager swarms is not currently guaranteed to succeed.
We recommend at least 3 managers, and if you have a lot of workers, you should pick 5 managers.
Enable if you want Docker for AWS to automatically cleanup unused space on your swarm nodes.
docker system prune will run staggered every day, starting at 1:42AM UTC on both workers and managers. The prune times are staggered slightly so that not all nodes will be pruned at the same time. This limits resource spikes on the swarm.
Pruning removes the following: - All stopped containers - All volumes not used by at least one container - All dangling images - All unused networks
Enable if you want Docker to send your container logs to CloudWatch. (“yes”, “no”) Defaults to yes.
Size of Workers’s ephemeral storage volume in GiB (20 - 1024).
Worker ephemeral storage volume type (“standard”, “gp2”).
Size of Manager’s ephemeral storage volume in GiB (20 - 1024)
Manager ephemeral storage volume type (“standard”, “gp2”)
Installing with the AWS Management Console
The simplest way to use the template is with the CloudFormation section of the AWS Management Console.
Go to the Release Notes page, and click on the “launch stack” button to start the deployment process.
Installing with the CLI
You can also invoke the Docker for AWS CloudFormation template from the AWS CLI:
Here is an example of how to use the CLI. Make sure you populate all of the parameters and their values:
$ aws cloudformation create-stack --stack-name teststack --template-url <templateurl> --parameters ParameterKey=KeyName,ParameterValue=<keyname> ParameterKey=InstanceType,ParameterValue=t2.micro ParameterKey=ManagerInstanceType,ParameterValue=t2.micro ParameterKey=ClusterSize,ParameterValue=1 --capabilities CAPABILITY_IAM
To fully automate installs, you can use the AWS Cloudformation API.
How it works
Docker for AWS starts with a CloudFormation template that will create everything that you need from scratch. There are only a few prerequisites that are listed above.
The CloudFormation template first creates a new VPC along with subnets and security groups. After the networking set-up completes, two Auto Scaling Groups are created, one for the managers and one for the workers, and the configured capacity setting is applied. Managers start first and create a quorum using Raft, then the workers start and join the swarm one at a time. At this point, the swarm is comprised of X number of managers and Y number of workers, and you can deploy your applications. See the deployment docs for your next steps.
If you increase the number of instances running in your worker Auto Scaling Group (via the AWS console, or updating the CloudFormation configuration), the new nodes that will start up will automatically join the swarm.
Elastic Load Balancers (ELBs) are set up to help with routing traffic to your swarm.
Docker for AWS automatically configures logging to Cloudwatch for containers you run on Docker for AWS. A Log Group is created for each Docker for AWS install, and a log stream for each container.
docker logs and
docker service logs commands are not supported on Docker for AWS when using Cloudwatch for logs. Instead, check container logs in CloudWatch.
Each node will have a few system containers running on them to help run your swarm cluster. In order for everything to run smoothly, please keep those containers running, and don’t make any changes. If you make any changes, Docker for AWS will not work correctly.