Scale ECS Services

Scale up or down any ECS service in a cluster without triggering a full deployment.

The ECS Scale step lets you scale any ECS service running in a cluster — independent of your deployment execution. Instead of triggering a full deployment stage to adjust instance counts, you can add this step to any pipeline and target any service in any ECS cluster you have access to.

This is useful when you need to scale services that are not part of the current deployment. For example, you might want to scale down a background worker service before deploying a new version of your primary application, or scale up a dependent service after deployment completes.

DELEGATE VERSION REQUIREMENT

This step requires a minimum delegate version of 88503.

If you are new to Harness ECS support, go to ECS deployments.

Key capabilities

• Scale any service in a cluster: Scale up or down any ECS service running in a cluster, even if that service is not part of the current deployment.

• Flexible infrastructure configuration: Optionally override the AWS connector, region, and cluster at the step level. If omitted, the step inherits these from the stage's infrastructure configuration.

• Works in Deploy and Custom stages: Use this step in a CD Deploy stage alongside other ECS steps, or in a Custom stage as a standalone utility.

Configure the ECS Scale step

1. In your pipeline, go to the Execution tab of your CD stage.

2. Select Add Step, then choose ECS Scale from the step library.

Click to view full size image

1. Configure the following fields:

Step parameters

Field	Description
Name	A name for the step.
Timeout	How long Harness should wait for the step to complete before failing. Default is 10m.
Connector	Reference to an AWS connector. Required or optional depending on the stage type (see below).
Region	The AWS region where the ECS cluster is running. Required or optional depending on the stage type (see below).
Cluster	The ECS cluster name. Required or optional depending on the stage type (see below).
Service	(Required) The name of the ECS service to scale. This value is always required and is not inherited from the stage's service or infrastructure configuration. You can select from the list of services running in the target cluster.
Instance Unit	Choose Count (absolute number of tasks) or Percentage (percentage of the service's current desired count).
Instances	The target number or percentage of instances to scale to.
Skip Steady State Check	When enabled, Harness triggers the scale operation but does not wait for the service to reach a steady state. Default is false.

Behavior in Deploy vs. Custom stages

The Connector, Region, and Cluster fields behave differently depending on the stage type:

• Deploy stage: These fields are optional. If left empty, the step inherits the values from the stage's infrastructure configuration. If you provide values in the step, they take precedence over the infrastructure configuration. This lets you target a service in a different cluster, region, or AWS account within the same Deploy stage.

• Custom stage: All fields are required. Since Custom stages do not have an infrastructure configuration to inherit from, you must explicitly provide the Connector, Region, and Cluster.

YAML example

ECS Scale step YAML

The following YAML scales the my_backend_service to 4 instances in a specific cluster:

- step:
    type: EcsScale
    name: Scale Backend Service
    identifier: ScaleBackendService
    spec:
      service: my_backend_service
      instanceUnit: Count
      instances: 4
      connectorRef: my_aws_connector
      region: us-east-1
      cluster: my-ecs-cluster
    timeout: 10m

Pipeline example

This pipeline shows a common use case: deploying a primary ECS service using a basic deployment strategy, then scaling a different backend service in a separate cluster after an approval gate. The ECS Scale step targets my_backend_service on my-ecs-cluster using its own AWS connector and region, independent of the infrastructure defined in the stage's environment.

This pattern is useful when a deployment to one service requires a dependent service to be scaled up (or down) as part of the same pipeline — without triggering a full redeployment of that dependent service.

Pipeline YAML with ECS Scale step

pipeline:
  name: ecs_basic_with_scale
  identifier: ecs_basic_with_scale
  projectIdentifier: my_project
  orgIdentifier: default
  tags: {}
  stages:
    - stage:
        name: ecs_deploy
        identifier: ecs_deploy
        description: ""
        type: Deployment
        spec:
          deploymentType: ECS
          service:
            serviceRef: my_ecs_service
          environment:
            environmentRef: staging
            deployToAll: false
            infrastructureDefinitions:
              - identifier: my_ecs_infra
          execution:
            steps:
              - stepGroup:
                  name: Basic Deployment
                  identifier: basicDeployment
                  steps:
                    - step:
                        name: ECS Service Setup
                        identifier: EcsServiceSetup
                        type: EcsServiceSetup
                        timeout: 10m
                        spec:
                          resizeStrategy: ResizeNewFirst
                    - step:
                        name: ECS Upgrade Container
                        identifier: EcsUpgradeContainer
                        type: EcsUpgradeContainer
                        timeout: 10m
                        spec:
                          newServiceInstanceCount: 5
                          newServiceInstanceUnit: Count
                          downsizeOldServiceInstanceCount: 1
                          downsizeOldServiceInstanceUnit: Count
              - step:
                  type: HarnessApproval
                  name: Approval Before Scale
                  identifier: ApprovalBeforeScale
                  spec:
                    approvalMessage: Please review and approve before scaling the backend service.
                    includePipelineExecutionHistory: true
                    isAutoRejectEnabled: false
                    approvers:
                      userGroups:
                        - account._account_all_users
                      minimumCount: 1
                      disallowPipelineExecutor: false
                    approverInputs: []
                  timeout: 1d
              - step:
                  type: EcsScale
                  name: Scale Backend Service
                  identifier: ScaleBackendService
                  spec:
                    service: my_backend_service
                    instanceUnit: Count
                    instances: 4
                    connectorRef: my_aws_connector
                    region: us-east-1
                    cluster: my-ecs-cluster
                  timeout: 10m
            rollbackSteps:
              - step:
                  name: ECS Basic Rollback
                  identifier: EcsBasicRollback
                  type: EcsBasicRollback
                  timeout: 10m
                  spec: {}
        tags: {}
        failureStrategies:
          - onFailure:
              errors:
                - AllErrors
              action:
                type: StageRollback

Advanced settings

In the Advanced tab, you can configure:

• Delegate Selector

• Conditional Execution

• Failure Strategy

• Looping Strategy

• Policy Enforcement