ECS任务场景下Step Functions中.sync的使用及任务执行完成判定咨询

Hey there, let’s tackle your two questions about AWS Step Functions and ECS task integration with .sync mode—super common use cases, so glad you’re digging into the details!

The .sync integration pattern in Step Functions is designed to wait for an asynchronous ECS task to complete before moving on to the next state in your workflow. Here’s how to implement it properly:

• Define your ECS task state with .sync: In your Step Functions state machine JSON, use the arn:aws:states:::ecs:runTask.sync resource type. This tells Step Functions to initiate the task and hold the workflow until the task finishes.
• Required parameters: You’ll need to specify the ECS cluster, task definition, launch type (Fargate or EC2), and any overrides (like container commands or environment variables). Here’s a minimal example:

{
  "StartAt": "RunECSTask",
  "States": {
    "RunECSTask": {
      "Type": "Task",
      "Resource": "arn:aws:states:::ecs:runTask.sync",
      "Parameters": {
        "Cluster": "your-ecs-cluster-name",
        "TaskDefinition": "your-task-definition:1",
        "LaunchType": "Fargate",
        "NetworkConfiguration": {
          "AwsvpcConfiguration": {
            "Subnets": ["subnet-123456"],
            "SecurityGroups": ["sg-123456"],
            "AssignPublicIp": "ENABLED"
          }
        }
      },
      "End": true
    }
  }
}

• Key considerations:
  • Ensure your Step Functions execution role has permissions for ecs:RunTask, ecs:DescribeTasks, and ecs:ListTasks (these are needed to launch and track the task).
  • If you’re using Fargate, make sure your task definition has the right CPU/memory allocations and network settings to run successfully.
  • You can add input/output paths to pass data between states (e.g., pass a job ID to the ECS task as an environment variable).

Under the hood, here’s the process Step Functions uses to track your ECS task until it finishes:

1. Initial task launch: Step Functions calls the ECS RunTask API to start your task. ECS returns a task ARN immediately, which Step Functions stores.
2. Polling ECS for task status: Step Functions then repeatedly calls the ECS DescribeTasks API (using the task ARN) to check the task’s state. By default, it polls every 60 seconds, but you can adjust this with the HeartbeatSeconds parameter if needed.
3. State evaluation: Step Functions waits until the task enters the STOPPED state. Once stopped, it checks:
   • The stopCode to understand why the task stopped (e.g., EssentialContainerExited means the main container finished its work).
   • The container’s exitCode: If the exit code is 0, Step Functions marks the state as successful and proceeds. If it’s non-zero, the state fails (you can handle this with Retry or Catch blocks in your state machine).
4. Timeout handling: You can set a TimeoutSeconds parameter in the state definition to prevent Step Functions from waiting indefinitely if the task gets stuck. If the timeout is reached, the state fails.

For example, here’s how you might add retry logic for transient ECS failures:

"RunECSTask": {
  "Type": "Task",
  "Resource": "arn:aws:states:::ecs:runTask.sync",
  "Parameters": { /* ... */ },
  "Retry": [
    {
      "ErrorEquals": ["ECS.AmazonECSException"],
      "IntervalSeconds": 10,
      "MaxAttempts": 3,
      "BackoffRate": 2.0
    }
  ],
  "Catch": [
    {
      "ErrorEquals": ["States.ALL"],
      "Next": "HandleTaskFailure"
    }
  ],
  "End": true
}

内容的提问来源于stack exchange，提问作者srinidhi sridharan