ECS Services and Scheduling

While task definitions tell ECS what to run, services tell ECS how many to run and how to keep them running. An ECS service is a long-running configuration that maintains a specified number of task instances, replaces failed tasks, and integrates with load balancers and auto-scaling. This lesson covers services, scheduling strategies, deployments, and auto-scaling.

---

What Is an ECS Service?

An ECS service ensures that a desired number of tasks are running at all times. If a task fails, crashes, or is stopped, the service scheduler automatically launches a replacement.

Service: my-web-app (desired count: 3)

  +--------+    +--------+    +--------+
  | Task 1 |    | Task 2 |    | Task 3 |
  | (v1.0) |    | (v1.0) |    | (v1.0) |
  +--------+    +--------+    +--------+
       ↑             ↑             ↑
       └─────────────┼─────────────┘
                     |
              Service Scheduler
              "Keep 3 running"

Key Service Properties

Property	Description
Desired count	The number of task instances the service should maintain
Task definition	The blueprint used to launch tasks (including the revision number)
Launch type / capacity provider	Where tasks run (Fargate, EC2, or a mix)
Deployment configuration	How updates are rolled out (minimum healthy, maximum percent)
Load balancer	Distribute traffic across task instances
Service discovery	Register tasks with AWS Cloud Map for DNS-based discovery

---

Creating a Service

Using the AWS CLI

aws ecs create-service \
  --cluster my-app-cluster \
  --service-name my-web-app \
  --task-definition my-web-app:3 \
  --desired-count 3 \
  --launch-type FARGATE \
  --network-configuration '{
    "awsvpcConfiguration": {
      "subnets": ["subnet-0123456789abcdef0", "subnet-0abcdef1234567890"],
      "securityGroups": ["sg-0123456789abcdef0"],
      "assignPublicIp": "ENABLED"
    }
  }' \
  --load-balancers '[{
    "targetGroupArn": "arn:aws:elasticloadbalancing:eu-west-2:123456789012:targetgroup/my-tg/1234567890",
    "containerName": "web",
    "containerPort": 3000
  }]'

---

Scheduling Strategies

ECS supports two service scheduler types:

Replica Strategy (Default)

The replica strategy maintains a specified number of tasks across your cluster. This is the most common approach for web applications and APIs.

Desired: 3 tasks
Strategy: REPLICA

  AZ-a              AZ-b              AZ-c
  +--------+        +--------+        +--------+
  | Task 1 |        | Task 2 |        | Task 3 |
  +--------+        +--------+        +--------+

ECS spreads tasks across Availability Zones by default for high availability.

Daemon Strategy

The daemon strategy runs exactly one task per EC2 instance in the cluster. This is useful for monitoring agents, log collectors, or other infrastructure tasks that should run on every host.

Note: Daemon scheduling is only available with the EC2 launch type, not Fargate.

---

Task Placement (EC2 Launch Type)

When using EC2 instances, ECS uses placement strategies and placement constraints to decide which instances run your tasks:

Placement Strategies

Strategy	Description
spread	Distribute tasks evenly across a specified attribute (e.g. attribute:ecs.availability-zone)
binpack	Pack tasks onto the fewest instances possible (optimises cost)
random	Place tasks randomly

You can combine strategies. For example, spread across AZs first, then binpack within each AZ:

"placementStrategy": [
  { "type": "spread", "field": "attribute:ecs.availability-zone" },
  { "type": "binpack", "field": "memory" }
]

Placement Constraints

Constraint	Description
distinctInstance	Each task runs on a different EC2 instance
memberOf	Tasks only run on instances matching an expression

"placementConstraints": [
  { "type": "memberOf", "expression": "attribute:ecs.instance-type =~ t3.*" }
]

---

Deployment Configuration

When you update a service (e.g. deploy a new image version), ECS performs a rolling update by default.

Rolling Update Parameters