配置CloudWatch告警检测ECS任务数异常时提示数据不足，求排查

我用以下Terraform代码配置CloudWatch告警，期望当ECS任务的期望数量大于运行数量时触发告警：

resource "aws_cloudwatch_metric_alarm" "ecs_task_count_alarm" {
  for_each            = toset(var.ecs_services) # 将列表转为集合以遍历唯一服务名
  alarm_name          = "ECS_Tasks_Running_Check_${each.key}"
  comparison_operator = "GreaterThanThreshold"
  evaluation_periods  = 1
  threshold           = 0
  alarm_description   = "当ECS服务${each.key}的DesiredTaskCount大于RunningTaskCount时触发告警。"

  # RunningTaskCount指标查询
  metric_query {
    id = "running_task_count"
    metric {
      metric_name = "RunningTaskCount"
      namespace   = "AWS/ECS"
      period      = 60 # 1分钟
      stat        = "Average"
      dimensions = {
        ClusterName = var.ecs_cluster_name
        ServiceName = each.key
      }
    }
  }

  # DesiredTaskCount指标查询
  metric_query {
    id = "desired_task_count"
    metric {
      metric_name = "DesiredTaskCount"
      namespace   = "AWS/ECS"
      period      = 60 # 1分钟
      stat        = "Average"
      dimensions = {
        ClusterName = var.ecs_cluster_name
        ServiceName = each.key
      }
    }
  }

  # 计算期望与运行任务数差值的数学表达式
  metric_query {
    id          = "number_of_tasks_we_want_but_arent_getting"
    expression  = "desired_task_count - running_task_count"
    label       = "期望与运行任务数的差值"
    return_data = true
  }

  alarm_actions             = [aws_sns_topic.sns_alarms.arn]
  insufficient_data_actions = [aws_sns_topic.sns_alarms.arn]
  ok_actions                = [aws_sns_topic.sns_alarms.arn]
}

但目前该告警提示数据不足，其中ecs_services为服务名称列表["foo", "bar"]，ecs_cluster_name为运行该服务的集群名称。请问我遗漏了什么配置？

导致数据不足的核心问题及修复方案如下：

• 统计指标类型错误：RunningTaskCount和DesiredTaskCount是瞬时计数型指标，代表某一时刻的任务数量，使用Average统计会导致数据失真（比如1分钟内数值无变化时平均值有效，但如果指标波动或上报点不足，就会出现无有效数据的情况）。应该替换为Sum或Maximum，这两种统计类型更适配这类离散计数指标：

  stat = "Sum" # 或 "Maximum"
  

• 维度大小写匹配检查：CloudWatch的维度是大小写敏感的，需确认var.ecs_cluster_name和each.key（服务名）与ECS控制台显示的集群、服务名称完全一致，包括大小写、特殊字符，哪怕一个字符不匹配都会找不到指标数据。

• 指标延迟问题：ECS的CloudWatch指标通常有1-5分钟的上报延迟，刚创建的告警或服务需要等待一段时间，待指标数据正常上报后再观察告警状态。

修改后的核心代码片段（仅调整stat字段）：

# RunningTaskCount指标查询
metric_query {
  id = "running_task_count"
  metric {
    metric_name = "RunningTaskCount"
    namespace   = "AWS/ECS"
    period      = 60
    stat        = "Sum" # 替换为Sum或Maximum
    dimensions = {
      ClusterName = var.ecs_cluster_name
      ServiceName = each.key
    }
  }
}

# DesiredTaskCount指标查询
metric_query {
  id = "desired_task_count"
  metric {
    metric_name = "DesiredTaskCount"
    namespace   = "AWS/ECS"
    period      = 60
    stat        = "Sum" # 替换为Sum或Maximum
    dimensions = {
      ClusterName = var.ecs_cluster_name
      ServiceName = each.key
    }
  }
}

内容的提问来源于stack exchange，提问作者Jordan Mackie