CORS redirect service

This is a template for a very simple HTTP(S) redirect service that replies to all request with a 301 to a fixed location and includes CORS headers.

Operations

Requirements

• AWS CLI, with configuration (at least credentials and region). You may use its config files or set the AWS environment variable, e.g. export AWS="aws --profile ..."
• docker
• jq

First deployment

1. Create a CFN stack using the cfn/erc.yaml template. This needs to be done only once; you can re-use the same stack for multiple related deployments (e.g. testing and prod).
2. Build the Docker image using the CFN stack name you chose in step 1:

# ./ops/build-docker-image.sh <erc-cfn-stack-name> [<docker-image-tag>]
# For example:
./ops/build-docker-image.sh my-fantastic-service prod

3. Push the image to ECR:

# ./ops/push-docker-image.sh <erc-cfn-stack-name> [<docker-image-tag>]
# For example:
./ops/push-docker-image.sh my-fantastic-service prod

4. Create a CFN stack using the cfn/redirector.yaml template.

Re-deploying the Docker image

If you need to update the image (after changing anything under docker/):

• Repeat steps 2 and 3 from above
• Force an ECS deployment using the CFN stack name you chose for cfn/redirector.yml in step 4:

  # ./opts/restart-service.sh <cfn-stack-name>
  # For example:
  ./ops/restart-service.sh my-fantastic-service-prod
  

Scaling

The service will autoscale horizontally based on the values provided to the CFN template parameters: containers (and instances) are spun up or down to maintain at least MinTPSCapacity regardless of traffic, and approximately TargetSpareTPSCapacity above traffic.

All parameters can be updated through a CFN stack update.

To scale vertically, change the InstanceType parameter, and adjust the value of ContainerTPS accordingly. The service is network-bound so container performance depends on the instance type. These values been verified through load tests:

• With instance type c6i.large, ContainerTPS should be 2000
• With instance type t2.medium, ContainerTPS should be 350

NOTE: Updating InstanceType will not affect running instances. If you are scaling down you will need to manually terminate them; make sure to temporarily increase MinTPSCapacity before doing so to avoid outages.