如何用AWS CDK关联已有LB与ECS服务创建CloudWatch仪表盘？

我正在使用AWS CDK创建CloudWatch指标、部件和仪表盘，但不清楚如何关联已有的负载均衡器（Load Balancer）、ECS服务等资源来创建这些CloudWatch资源。以下是我的POC代码：

from aws_cdk import (
    Stack,
    aws_elasticloadbalancingv2 as elbv2,
    aws_cloudwatch as cw
)
from constructs import Construct

class CwDashboardStack(Stack):

    def __init__(self, scope: Construct, construct_id: str, **kwargs) -> None:
        super().__init__(scope, construct_id, **kwargs)

        cpu_utilization_metric = cw.Metric(namespace="AWS/ECS",metric_name="CPUUtilization")
        cpu_widget = cw.GraphWidget(
            title="CPU Utilization",
            height=8,
            width=12,
            left=[cpu_utilization_metric]
        )            

        cw.Dashboard(
            self,
            "Dashboard",
            dashboard_name="Service-Status",
            widgets=[
                [cpu_widget]
            ]
        )

但生成的CloudWatch仪表盘为空，这是因为未指定ECS服务名、集群等信息所致。但我找不到Metric类中可配置这些信息的对应属性，请问有解决建议吗？

核心解决思路 #

CloudWatch指标的资源关联逻辑通过Metric类的dimensions_map参数实现，需要为指标指定对应服务的维度信息（如ECS的集群名、服务名）。以下是针对ECS服务、负载均衡器的具体实现：

1. 关联已有ECS服务 #

ECS的CPU利用率指标必须指定ClusterName和ServiceName两个维度，可通过两种方式获取已有资源信息：

方式一：通过资源查找获取已有ECS集群和服务

from aws_cdk import aws_ecs as ecs

# 查找已有ECS集群
existing_cluster = ecs.Cluster.from_cluster_attributes(
    self, "ExistingCluster",
    cluster_name="your-cluster-name"
)

# 查找已有ECS服务
existing_service = ecs.FargateService.from_fargate_service_attributes(
    self, "ExistingService",
    cluster=existing_cluster,
    service_name="your-service-name"
)

# 创建带维度的CPU指标
cpu_utilization_metric = cw.Metric(
    namespace="AWS/ECS",
    metric_name="CPUUtilization",
    dimensions_map={
        "ClusterName": existing_cluster.cluster_name,
        "ServiceName": existing_service.service_name
    },
    period=Duration.minutes(1)
)

方式二：直接传入已知的集群和服务名称

若明确知道集群和服务名称，可直接硬编码维度值：

cpu_utilization_metric = cw.Metric(
    namespace="AWS/ECS",
    metric_name="CPUUtilization",
    dimensions_map={
        "ClusterName": "my-production-cluster",
        "ServiceName": "my-api-service"
    },
    statistic="Average"  # 指定统计类型，如Average、Sum、Maximum等
)

2. 关联已有负载均衡器（ALB） #

以ALB的请求计数指标为例，需指定LoadBalancer和TargetGroup维度：

# 查找已有ALB
existing_alb = elbv2.ApplicationLoadBalancer.from_load_balancer_attributes(
    self, "ExistingALB",
    load_balancer_arn="arn:aws:elasticloadbalancing:us-east-1:123456789012:loadbalancer/app/my-alb/abc123"
)

# 查找已有目标组
existing_tg = elbv2.ApplicationTargetGroup.from_target_group_attributes(
    self, "ExistingTG",
    target_group_arn="arn:aws:elasticloadbalancing:us-east-1:123456789012:targetgroup/my-tg/def456"
)

# 创建ALB请求计数指标
alb_request_metric = cw.Metric(
    namespace="AWS/ApplicationELB",
    metric_name="RequestCount",
    dimensions_map={
        "LoadBalancer": existing_alb.load_balancer_name,
        "TargetGroup": existing_tg.target_group_name
    }
)

3. 完整仪表盘示例代码 #

from aws_cdk import (
    Stack,
    aws_elasticloadbalancingv2 as elbv2,
    aws_cloudwatch as cw,
    aws_ecs as ecs,
    Duration
)
from constructs import Construct

class CwDashboardStack(Stack):

    def __init__(self, scope: Construct, construct_id: str, **kwargs) -> None:
        super().__init__(scope, construct_id, **kwargs)

        # 关联已有ECS服务
        existing_cluster = ecs.Cluster.from_cluster_attributes(
            self, "ExistingCluster",
            cluster_name="your-cluster-name"
        )
        existing_service = ecs.FargateService.from_fargate_service_attributes(
            self, "ExistingService",
            cluster=existing_cluster,
            service_name="your-service-name"
        )

        cpu_utilization_metric = cw.Metric(
            namespace="AWS/ECS",
            metric_name="CPUUtilization",
            dimensions_map={
                "ClusterName": existing_cluster.cluster_name,
                "ServiceName": existing_service.service_name
            },
            period=Duration.minutes(1),
            statistic="Average"
        )

        # 关联已有ALB
        existing_alb = elbv2.ApplicationLoadBalancer.from_load_balancer_attributes(
            self, "ExistingALB",
            load_balancer_arn="your-alb-arn"
        )
        existing_tg = elbv2.ApplicationTargetGroup.from_target_group_attributes(
            self, "ExistingTG",
            target_group_arn="your-tg-arn"
        )

        alb_request_metric = cw.Metric(
            namespace="AWS/ApplicationELB",
            metric_name="RequestCount",
            dimensions_map={
                "LoadBalancer": existing_alb.load_balancer_name,
                "TargetGroup": existing_tg.target_group_name
            },
            period=Duration.minutes(1)
        )

        # 创建仪表盘部件
        cpu_widget = cw.GraphWidget(
            title="ECS CPU Utilization",
            height=8,
            width=12,
            left=[cpu_utilization_metric]
        )

        alb_widget = cw.GraphWidget(
            title="ALB Request Count",
            height=8,
            width=12,
            left=[alb_request_metric]
        )

        # 部署仪表盘
        cw.Dashboard(
            self,
            "ServiceStatusDashboard",
            dashboard_name="Service-Status",
            widgets=[
                [cpu_widget, alb_widget]
            ]
        )

关键注意事项 #

• 维度名称必须与CloudWatch对应服务的指标维度完全一致（可在AWS控制台CloudWatch指标页面确认）。
• 跨CDK栈的资源建议通过CfnOutput传递ARN或名称，避免硬编码。
• 根据监控需求选择合适的statistic参数值，如Average用于查看资源平均负载，Sum用于统计请求总量。

内容的提问来源于stack exchange，提问作者shikha singh