Fork me on GitHub

How to configure an AWS Global Accelerator in CloudFormation

AWS Global Accelerator is an awesome service which was announced in 2018, but unfortunately without CloudFormation support. In this blog we show you how to configure your accelerators, listeners and endpoint groups in CloudFormation using a custom provider.

How do I use it?

Very simple, add a custom global accelerator, listener and one or more endpoint groups to your template as shown below:

    Type: Custom::GlobalAccelerator
      Name: my-global-accelerator
      Enabled: true
      IpAddressType: IPV4
      ServiceToken: !Sub 'arn:aws:lambda:${AWS::Region}:${AWS::AccountId}:function:cfn-global-accelerator-provider'

    Type: Custom::GlobalAcceleratorListener
      AcceleratorArn: !Ref Accelerator
      Protocol: TCP
        - FromPort: 80
          ToPort: 80
      ClientAffinity: NONE
      ServiceToken: !Sub 'arn:aws:lambda:${AWS::Region}:${AWS::AccountId}:function:cfn-global-accelerator-provider'

Using the Listener ARN, you can now add endpoint groups to point to your regional resources:

    Type: Custom::GlobalAcceleratorEndpointGroup
      ServiceToken: !Sub 'arn:aws:lambda:${AWS::Region}:${AWS::AccountId}:function:cfn-global-accelerator-provider'
      ListenerArn: !Ref GlobalAcceleratorListener
      EndpointGroupRegion: !Ref AWS::Region
        - EndpointId: !Ref LoadBalancer
          Weight: 100

    Type: AWS::ElasticLoadBalancingV2::LoadBalancer
      Subnets: !Ref PublicSubnets
      - !Ref ALBSecurityGroup

Deploy the provider

To deploy the provider, type:

git clone https://github.com/binxio/cfn-global-accelerator-provider.git
cd cfn-global-accelerator-provider

aws --region us-west-2 cloudformation create-stack \
        --capabilities CAPABILITY_IAM \
        --stack-name cfn-global-accelerator-provider \
        --template-body file://./cloudformation/templates/provider.yaml

aws --region us-west-2 cloudformation wait stack-create-complete  --stack-name cfn-global-accelerator-provider

This CloudFormation template will use our pre-packaged provider from s3://binxio-public/lambdas/cfn-global-accelerator-provider-0.1.1.zip.


The following tree shows the deployment configuration of our global accelerator demo:

└── us-west-2
    ├── accelerator.yaml
│   ├── provider.yaml
├── eu-central-1
│   ├── app.yaml
│   ├── provider.yaml
│   └── vpc.yaml
├── eu-west-1
│   ├── app.yaml
│   ├── provider.yaml
│   └── vpc.yaml

The accelerator is deployed in us-west-2 and the application is deployed in both eu-central-1 and eu-west-1 respectively. The custom provider is also deployed in the application regions in order to create the global accelerator endpoint groups.

Deploy the demo

In order to deploy the demo, type:

cd cloudformation
pip install sceptre
sceptre launch -y demo

to view the application, type:

  aws --region us-west-2 globalaccelerator \
  list-accelerators \
    --output text \
    --query 'Accelerators[?Name==`cfn-global-accelerator-demo-us-west-2-accelerator`].IpSets[0].IpAddresses[0]')

open http://$GA_IP_ADDRESS


This provider was created within a relaxed working day and allows you to configure your accelerators, listeners and endpoint groups in CloudFormation. Although creating this temporary custom provider cost less than a day, it is not the ideal solution. It has to be created, installed and decommissioned when Amazon finally provides support for the Global Accelerator resources.

CloudFormation’s open-source rival terraform provided support for the Global Accelerator before AWS, as shown in Martijn’s blog on building a global application with AWS Global Accelerator. If it helps, I would like to offer AWS our services to ensure that CloudFormation is always on-par with the API before the feature is released.

Picture of Mark van Holsteijn
Mark van Holsteijn