Updating a CodeDeploy-managed ECS Service from the AWS Console resets the capacity provider to FARGATE

I had a FARGATE setup where the capacity provider strategy put all the weight on FARGATE_SPOT with weight = 1.

• FARGATE_SPOT weight = 1
• FARGATE weight = 0

The problem when updating an ECS Service from the AWS Console

In Configure service, I had set FARGATE_SPOT weight = 1.

For a quick test, I made a small change to the task definition and deployed it by updating only the task from the AWS Console.

And then…

The capacity provider strategy configuration was gone…

Why does this happen?

When you go through the ECS Service update steps in the AWS Console and click the AppSpec for the Revision under Configure CodeDeploy Deployment, you find the following configuration:

{
    "version": 1,
    "Resources": [
        {
            "TargetService": {
                "Type": "AWS::ECS::Service",
                "Properties": {
                    "TaskDefinition": "<タスク定義 ARN>",
                    "LoadBalancerInfo": {
                        "ContainerName": "nginx",
                        "ContainerPort": 80
                    },
                    "PlatformVersion": "LATEST"
                }
            }
        }
    ]
}

There is no capacity provider strategy configuration here.

As a result, after the update completed, the capacity provider strategy configuration was gone.

If you manage the capacity provider strategy with Terraform, running terraform apply will unintentionally force the ECS Service to be recreated, causing downtime.

  # aws_ecs_service.app must be replaced
-/+ resource "aws_ecs_service" "app" {

      + capacity_provider_strategy { # forces replacement
          + capacity_provider = "FARGATE"
          + weight            = 0
        }
      + capacity_provider_strategy { # forces replacement
          + capacity_provider = "FARGATE_SPOT"
          + weight            = 1
        }

Solution

For an ECS Service managed by CodeDeploy, perform updates via the AWS CLI or through CodeDeploy.

Trying to handle everything with options to aws deploy create-deployment gets a bit complicated.

aws deploy create-deployment \
	--application-name xxx \
	--deployment-group-name yyy \
	--revision '{"revisionType": "AppSpecContent", "appSpecContent": {"content": "{\"version\": 0.0, \"Resources\": [{\"TargetService\": {\"Type\": \"AWS::ECS::Service\", \"Properties\": {\"TaskDefinition\": \"<タスク定義 ARN>\", \"LoadBalancerInfo\": {\"ContainerName\": \"nginx\", \"ContainerPort\": 80},\"CapacityProviderStrategy\": [{\"Base\":0,\"CapacityProvider\":\"FARGATE_SPOT\",\"Weight\":1},{\"Base\":0,\"CapacityProvider\":\"FARGATE\",\"Weight\":0}]}}}]}"}}'

Putting it into a file keeps the command simple ♪
(Since the content definition is a String, it has to be written as a single-line string, which makes it a bit hard to read.)

aws deploy create-deployment --cli-input-json file://deployment.json

deployment.json

{
    "applicationName": "xxx",
    "deploymentGroupName": "yyy",
    "revision": {
	    "revisionType": "AppSpecContent",
		  "appSpecContent": {
			  "content": "{\"version\": 0.0, \"Resources\": [{\"TargetService\": {\"Type\": \"AWS::ECS::Service\", \"Properties\": {\"TaskDefinition\": \"<タスク定義 ARN>\", \"LoadBalancerInfo\": {\"ContainerName\": \"nginx\", \"ContainerPort\": 80},\"CapacityProviderStrategy\": [{\"Base\":0,\"CapacityProvider\":\"FARGATE_SPOT\",\"Weight\":1},{\"Base\":0,\"CapacityProvider\":\"FARGATE\",\"Weight\":0}]}}}]}"
		}
	}
}

I confirmed that by explicitly specifying the appspec, you can explicitly preserve the capacity provider strategy.

That’s all.
I hope this is helpful.