如何通过AWS Step Functions自身逻辑判断执行状态并控制依赖执行？

问题背景 #

我们每日导入两个数据集（dataset A和dataset B），其中dataset A可能并非每日都会出现；这两个数据集分别触发独立的AWS Step Functions；处理dataset B的Step Function B必须等到处理dataset A的Step Function A停止运行或从未运行后才能执行，以确保Step Function B使用最新数据。

当前调用ListExecutions获取的执行历史输出如下：

{
  "executions": [
        {
            "executionArn": "arn:aws:states:us-west-2:123456789:execution:StateMachineA12345-12345-12345:1234-1",
            "stateMachineArn": "arn:aws:states:us-west-2:123456789:stateMachine:StateMachineA12345-12345-12345",
            "name": "1234-1",
            "status": "RUNNING",
            "startDate": "2023-02-27T15:21:22.205000-08:00",
            "stopDate": "2023-02-27T15:28:11.358000-08:00"
        },
        {
            "executionArn": "arn:aws:states:us-west-2:123456789:execution:StateMachineA12345-12345-12345:1234-2",
            "stateMachineArn": "arn:aws:states:us-west-2:123456789:stateMachine:StateMachineA12345-12345-12345",
            "name": "1234-2",
            "status": "ABORTED",
            "startDate": "2023-02-27T15:19:55.739000-08:00",
            "stopDate": "2023-02-27T15:21:11.924000-08:00"
        },
        {
            "executionArn": "arn:aws:states:us-west-2:123456789:execution:StateMachineA12345-12345-12345:1234-3",
            "stateMachineArn": "arn:aws:states:us-west-2:123456789:stateMachine:StateMachineA12345-12345-12345",
            "name": "1234-3",
            "status": "COMPLETED",
            "startDate": "2023-02-27T15:18:45.228000-08:00",
            "stopDate": "2023-02-27T15:19:20.651000-08:00"
        },
        {
            "executionArn": "arn:aws:states:us-west-2:123456789:execution:StateMachineA12345-12345-12345:1234-4",
            "stateMachineArn": "arn:aws:states:us-west-2:123456789:stateMachine:StateMachineA12345-12345-12345",
            "name": "1234-4",
            "status": "FAILED",
            "startDate": "2023-02-27T15:18:30.145000-08:00",
            "stopDate": "2023-02-27T15:18:34.315000-08:00"
        }
    ]
}

需要解决两个问题：

1. 能否仅通过Step Functions逻辑，添加ListExecutions步骤并等待直到State Machine A的执行状态不等于RUNNING？
2. 能否仅通过Step Functions逻辑，将上述ListExecutions的输出转换为{"status": "RUNNING"}或{"status": "NOT RUNNING"}结构？

问题1：实现等待State Machine A完成或未运行的逻辑 #

完全可以通过Step Functions内置状态实现，无需Lambda，核心流程如下：

核心步骤 #

• 调用ListExecutions时过滤运行中执行：指定State Machine A的ARN，设置StatusFilter: "RUNNING"，直接返回正在运行的执行（如果存在），避免处理全量历史数据。
• 循环检查直到无运行中执行：用Choice状态判断返回的executions数组长度：
  • 数组长度>0：进入Wait状态（比如等待30秒），然后回到ListExecutions步骤重复检查。
  • 数组长度=0（包括从未运行过的情况）：直接执行State Machine B的后续逻辑。

示例状态机片段 #

{
  "States": {
    "CheckSMARunningStatus": {
      "Type": "Task",
      "Resource": "arn:aws:states:::aws-sdk:states:listExecutions",
      "Parameters": {
        "StateMachineArn": "arn:aws:states:us-west-2:123456789:stateMachine:StateMachineA12345-12345-12345",
        "StatusFilter": "RUNNING"
      },
      "Next": "HasRunningExec?"
    },
    "HasRunningExec?": {
      "Type": "Choice",
      "Choices": [
        {
          "Variable": "$.executions.length()",
          "NumericGreaterThan": 0,
          "Next": "Wait30Seconds"
        }
      ],
      "Default": "StartSMBProcessing"
    },
    "Wait30Seconds": {
      "Type": "Wait",
      "Seconds": 30,
      "Next": "CheckSMARunningStatus"
    },
    "StartSMBProcessing": {
      "Type": "Pass",
      "End": true
      // 替换为State Machine B的实际处理逻辑
    }
  }
}

问题2：转换ListExecutions输出为指定状态结构 #

同样无需Lambda，通过Step Functions的Task、Choice和Pass状态即可实现，推荐两种方法：

方法1：基于过滤后的运行中执行判断（高效优先） #

直接利用ListExecutions的StatusFilter参数减少数据量，再判断输出：

1. 调用ListExecutions筛选出State Machine A的所有RUNNING执行。
2. 用Choice状态检查executions数组长度：
   • 长度>0：通过Pass状态输出{"status": "RUNNING"}。
   • 长度=0：通过Pass状态输出{"status": "NOT RUNNING"}。

示例状态机片段 #

{
  "States": {
    "FetchRunningExecs": {
      "Type": "Task",
      "Resource": "arn:aws:states:::aws-sdk:states:listExecutions",
      "Parameters": {
        "StateMachineArn": "arn:aws:states:us-west-2:123456789:stateMachine:StateMachineA12345-12345-12345",
        "StatusFilter": "RUNNING"
      },
      "Next": "ResolveStatus"
    },
    "ResolveStatus": {
      "Type": "Choice",
      "Choices": [
        {
          "Variable": "$.executions.length()",
          "NumericGreaterThan": 0,
          "Next": "ReturnRunningStatus"
        }
      ],
      "Default": "ReturnNotRunningStatus"
    },
    "ReturnRunningStatus": {
      "Type": "Pass",
      "Result": {
        "status": "RUNNING"
      },
      "ResultPath": "$",
      "End": true
    },
    "ReturnNotRunningStatus": {
      "Type": "Pass",
      "Result": {
        "status": "NOT RUNNING"
      },
      "ResultPath": "$",
      "End": true
    }
  }
}

方法2：处理全量执行列表（适用于必须获取全量历史的场景） #

如果无法使用StatusFilter，可以通过Map和Filter状态提取并筛选状态：

1. 用Map遍历全量executions数组，提取每个执行的status字段。
2. 用Filter状态筛选出值为RUNNING的条目。
3. 判断筛选后的数组长度，输出对应状态结构。

这种方法步骤更多，性能不如方法1，仅在特殊场景下使用。

内容的提问来源于stack exchange，提问作者Justin Bates