CloudWatch Alarms

最終更新: February 20, 2026

CloudWatch Alarms の概要

Amazon CloudWatch Alarms は、AWS ネイティブモニタリングサービスのアラート機能です。CloudWatch Metrics に基づいてアラートを作成し、SNS、Lambda、EC2 Auto Scaling などとの統合を通じて自動応答を可能にします。

主な機能

Metric Alarms: 単一のメトリクスに基づくアラート
Composite Alarms: 複数のアラーム条件を組み合わせる
異常検知: 機械学習ベースの異常検知
Alarm Actions: アラート発生時に自動アクションを実行
AWS Service Integration: EC2、ECS、EKS、Lambda などとのネイティブ統合

CloudWatch Alarms と Prometheus Alertmanager の比較

特性	CloudWatch Alarms	Prometheus Alertmanager
タイプ	AWS Managed Service	Open Source
データソース	CloudWatch Metrics	Prometheus Metrics
クエリ言語	CloudWatch Metrics Math	PromQL
コスト	アラームごとの料金	無料（インフラストラクチャコストのみ）
複雑なルーティング	制限あり	高度なルーティングをサポート
AWS 統合	ネイティブ	追加設定が必要

アーキテクチャ

CloudWatch Alarms の動作フロー

アラームの状態

CloudWatch Alarms には 3 つの状態があります。

Metric Alarms

基本的なアラーム作成（Console/CLI）

AWS CLI

bash

# Create CPU utilization alarm
aws cloudwatch put-metric-alarm \
  --alarm-name "HighCPUUtilization" \
  --alarm-description "CPU usage exceeds 80%" \
  --metric-name CPUUtilization \
  --namespace AWS/EC2 \
  --statistic Average \
  --period 300 \
  --threshold 80 \
  --comparison-operator GreaterThanThreshold \
  --evaluation-periods 2 \
  --dimensions Name=InstanceId,Value=i-1234567890abcdef0 \
  --alarm-actions arn:aws:sns:ap-northeast-2:123456789012:alerts \
  --ok-actions arn:aws:sns:ap-northeast-2:123456789012:alerts \
  --treat-missing-data notBreaching

アラーム設定の構成要素

パラメータ	説明	例
`metric-name`	監視するメトリクスの名前	`CPUUtilization`
`namespace`	メトリクスの名前空間	`AWS/EC2`, `AWS/EKS`
`statistic`	統計関数	`Average`, `Sum`, `Maximum`, `Minimum`, `p99`
`period`	評価期間（秒）	`60`, `300`, `3600`
`threshold`	しきい値	`80`
`comparison-operator`	比較演算子	`GreaterThanThreshold`
`evaluation-periods`	連続評価回数	`2`（2 回連続で超過した場合にアラート）
`datapoints-to-alarm`	アラームに必要なデータポイント	`3` 件中 `2` 件
`treat-missing-data`	欠損データの処理	`notBreaching`, `breaching`, `ignore`, `missing`

比較演算子

yaml

# Available comparison operators
comparison-operators:
  - GreaterThanThreshold           # Greater than
  - GreaterThanOrEqualToThreshold  # Greater than or equal
  - LessThanThreshold              # Less than
  - LessThanOrEqualToThreshold     # Less than or equal
  - LessThanLowerOrGreaterThanUpperThreshold  # Outside range
  - LessThanLowerThreshold         # Below lower bound
  - GreaterThanUpperThreshold      # Above upper bound

Metrics Math を使用するアラーム

bash

# Error rate calculation alarm (error count / total requests)
aws cloudwatch put-metric-alarm \
  --alarm-name "HighErrorRate" \
  --alarm-description "Error rate exceeds 5%" \
  --metrics '[
    {
      "Id": "errors",
      "MetricStat": {
        "Metric": {
          "Namespace": "AWS/ApplicationELB",
          "MetricName": "HTTPCode_Target_5XX_Count",
          "Dimensions": [
            {"Name": "LoadBalancer", "Value": "app/my-alb/1234567890"}
          ]
        },
        "Period": 300,
        "Stat": "Sum"
      },
      "ReturnData": false
    },
    {
      "Id": "requests",
      "MetricStat": {
        "Metric": {
          "Namespace": "AWS/ApplicationELB",
          "MetricName": "RequestCount",
          "Dimensions": [
            {"Name": "LoadBalancer", "Value": "app/my-alb/1234567890"}
          ]
        },
        "Period": 300,
        "Stat": "Sum"
      },
      "ReturnData": false
    },
    {
      "Id": "error_rate",
      "Expression": "(errors / requests) * 100",
      "ReturnData": true
    }
  ]' \
  --threshold 5 \
  --comparison-operator GreaterThanThreshold \
  --evaluation-periods 2 \
  --alarm-actions arn:aws:sns:ap-northeast-2:123456789012:alerts

Metrics Math 関数

yaml

# Commonly used functions
math-functions:
  # Arithmetic operations
  - "m1 + m2"           # Sum
  - "m1 - m2"           # Difference
  - "m1 * m2"           # Product
  - "m1 / m2"           # Division
  - "(m1 / m2) * 100"   # Percentage

  # Statistical functions
  - "AVG(METRICS())"    # Average
  - "SUM(METRICS())"    # Sum
  - "MIN(METRICS())"    # Minimum
  - "MAX(METRICS())"    # Maximum

  # Conditional functions
  - "IF(m1 > 100, m1, 0)"  # Conditional

  # Time-related
  - "RATE(m1)"          # Rate of change
  - "DIFF(m1)"          # Difference
  - "PERIOD(m1)"        # Period

  # Search
  - "SEARCH('{AWS/EC2,InstanceId} MetricName=\"CPUUtilization\"', 'Average', 300)"

Composite Alarms

Composite Alarm の概念

Composite Alarms では、複数の Metric Alarms を組み合わせて複雑な条件を定義できます。

Composite Alarms の作成

bash

# Create individual alarms
aws cloudwatch put-metric-alarm \
  --alarm-name "HighCPU" \
  --metric-name CPUUtilization \
  --namespace AWS/EC2 \
  --statistic Average \
  --period 300 \
  --threshold 80 \
  --comparison-operator GreaterThanThreshold \
  --evaluation-periods 2 \
  --dimensions Name=InstanceId,Value=i-1234567890abcdef0

aws cloudwatch put-metric-alarm \
  --alarm-name "HighMemory" \
  --metric-name mem_used_percent \
  --namespace CWAgent \
  --statistic Average \
  --period 300 \
  --threshold 85 \
  --comparison-operator GreaterThanThreshold \
  --evaluation-periods 2 \
  --dimensions Name=InstanceId,Value=i-1234567890abcdef0

aws cloudwatch put-metric-alarm \
  --alarm-name "HighDisk" \
  --metric-name disk_used_percent \
  --namespace CWAgent \
  --statistic Average \
  --period 300 \
  --threshold 90 \
  --comparison-operator GreaterThanThreshold \
  --evaluation-periods 2 \
  --dimensions Name=InstanceId,Value=i-1234567890abcdef0

# Create Composite Alarm
aws cloudwatch put-composite-alarm \
  --alarm-name "ServerResourceCritical" \
  --alarm-description "Server resources are critical" \
  --alarm-rule "ALARM(HighCPU) AND ALARM(HighMemory) OR ALARM(HighDisk)" \
  --alarm-actions arn:aws:sns:ap-northeast-2:123456789012:critical-alerts \
  --ok-actions arn:aws:sns:ap-northeast-2:123456789012:alerts

アラームルールの構文

yaml

# Composite Alarm rule syntax
rule-syntax:
  # Basic operators
  - "ALARM(alarm-name)"      # Check ALARM state
  - "OK(alarm-name)"         # Check OK state
  - "INSUFFICIENT_DATA(alarm-name)"  # Check INSUFFICIENT_DATA state

  # Logical operators
  - "AND"                    # All conditions met
  - "OR"                     # One or more conditions met
  - "NOT"                    # Negation
  - "()"                     # Grouping

examples:
  # All conditions met
  - "ALARM(A1) AND ALARM(A2) AND ALARM(A3)"

  # One or more met
  - "ALARM(A1) OR ALARM(A2)"

  # Complex condition
  - "(ALARM(A1) AND ALARM(A2)) OR ALARM(A3)"

  # Negation
  - "ALARM(A1) AND NOT ALARM(A2)"

  # M of N pattern (2 or more of 3)
  - "(ALARM(A1) AND ALARM(A2)) OR (ALARM(A1) AND ALARM(A3)) OR (ALARM(A2) AND ALARM(A3))"

アラート抑制パターン

bash

# Suppress alerts during maintenance
aws cloudwatch put-composite-alarm \
  --alarm-name "ProductionAlerts" \
  --alarm-rule "ALARM(HighCPU) AND NOT ALARM(MaintenanceMode)" \
  --alarm-actions arn:aws:sns:ap-northeast-2:123456789012:alerts

# Manually transition MaintenanceMode alarm to ALARM state for suppression
aws cloudwatch set-alarm-state \
  --alarm-name "MaintenanceMode" \
  --state-value ALARM \
  --state-reason "Scheduled maintenance"

異常検知

異常検知の概要

CloudWatch Anomaly Detection は機械学習を使用してメトリクスの通常パターンを学習し、外れ値を検出します。

異常検知アラームの作成

bash

# Anomaly Detection model creation (automatic)
# Model is automatically created when first alarm is created

aws cloudwatch put-metric-alarm \
  --alarm-name "CPUAnomalyDetection" \
  --alarm-description "CPU usage is anomalous" \
  --metrics '[
    {
      "Id": "m1",
      "MetricStat": {
        "Metric": {
          "Namespace": "AWS/EC2",
          "MetricName": "CPUUtilization",
          "Dimensions": [
            {"Name": "InstanceId", "Value": "i-1234567890abcdef0"}
          ]
        },
        "Period": 300,
        "Stat": "Average"
      },
      "ReturnData": true
    },
    {
      "Id": "ad1",
      "Expression": "ANOMALY_DETECTION_BAND(m1, 2)",
      "ReturnData": true
    }
  ]' \
  --threshold-metric-id ad1 \
  --comparison-operator LessThanLowerOrGreaterThanUpperThreshold \
  --evaluation-periods 2 \
  --alarm-actions arn:aws:sns:ap-northeast-2:123456789012:alerts

異常検知の設定

yaml

# ANOMALY_DETECTION_BAND function
# ANOMALY_DETECTION_BAND(metric, stddev)
# - metric: Metric to analyze
# - stddev: Standard deviation multiplier (default 2)

examples:
  # 2 standard deviations (approximately 95% confidence interval)
  - "ANOMALY_DETECTION_BAND(m1, 2)"

  # 3 standard deviations (approximately 99.7% confidence interval)
  - "ANOMALY_DETECTION_BAND(m1, 3)"

  # More sensitive detection (1 standard deviation)
  - "ANOMALY_DETECTION_BAND(m1, 1)"

モデル学習期間の調整

bash

# Add exclusion periods to existing model (maintenance, incident periods, etc.)
aws cloudwatch put-anomaly-detector \
  --namespace AWS/EC2 \
  --metric-name CPUUtilization \
  --stat Average \
  --dimensions Name=InstanceId,Value=i-1234567890abcdef0 \
  --configuration '{
    "ExcludedTimeRanges": [
      {
        "StartTime": "2025-02-15T00:00:00Z",
        "EndTime": "2025-02-15T06:00:00Z"
      }
    ]
  }'

bash

# Create SNS Topic
aws sns create-topic --name eks-alerts

# Add Email subscription
aws sns subscribe \
  --topic-arn arn:aws:sns:ap-northeast-2:123456789012:eks-alerts \
  --protocol email \
  --notification-endpoint team@example.com

# Add SMS subscription
aws sns subscribe \
  --topic-arn arn:aws:sns:ap-northeast-2:123456789012:eks-alerts \
  --protocol sms \
  --notification-endpoint +821012345678

# Add Lambda subscription
aws sns subscribe \
  --topic-arn arn:aws:sns:ap-northeast-2:123456789012:eks-alerts \
  --protocol lambda \
  --notification-endpoint arn:aws:lambda:ap-northeast-2:123456789012:function:alert-handler

json

// Subscription filter policy
{
  "severity": ["critical", "high"],
  "environment": ["production"]
}

bash

# Apply filter policy
aws sns set-subscription-attributes \
  --subscription-arn arn:aws:sns:ap-northeast-2:123456789012:eks-alerts:xxx \
  --attribute-name FilterPolicy \
  --attribute-value '{"severity": ["critical", "high"]}'

python

# lambda_function.py
import json
import urllib3
import os

http = urllib3.PoolManager()

def lambda_handler(event, context):
    slack_webhook_url = os.environ['SLACK_WEBHOOK_URL']

    for record in event['Records']:
        sns_message = json.loads(record['Sns']['Message'])

        # Parse CloudWatch Alarm message
        alarm_name = sns_message.get('AlarmName', 'Unknown')
        alarm_description = sns_message.get('AlarmDescription', '')
        new_state = sns_message.get('NewStateValue', 'Unknown')
        reason = sns_message.get('NewStateReason', '')
        timestamp = sns_message.get('StateChangeTime', '')

        # Slack message color
        if new_state == 'ALARM':
            color = '#ff0000'
            emoji = ':rotating_light:'
        elif new_state == 'OK':
            color = '#36a64f'
            emoji = ':white_check_mark:'
        else:
            color = '#808080'
            emoji = ':question:'

        # Compose Slack message
        slack_message = {
            "attachments": [
                {
                    "color": color,
                    "title": f"{emoji} {alarm_name}",
                    "text": alarm_description,
                    "fields": [
                        {
                            "title": "State",
                            "value": new_state,
                            "short": True
                        },
                        {
                            "title": "Time",
                            "value": timestamp,
                            "short": True
                        },
                        {
                            "title": "Reason",
                            "value": reason,
                            "short": False
                        }
                    ]
                }
            ]
        }

        # Send to Slack
        response = http.request(
            'POST',
            slack_webhook_url,
            body=json.dumps(slack_message),
            headers={'Content-Type': 'application/json'}
        )

    return {'statusCode': 200}

EventBridge 統合

EventBridge ルールの作成

bash

# Route CloudWatch Alarm state changes to EventBridge
aws events put-rule \
  --name "CloudWatchAlarmStateChange" \
  --event-pattern '{
    "source": ["aws.cloudwatch"],
    "detail-type": ["CloudWatch Alarm State Change"],
    "detail": {
      "state": {
        "value": ["ALARM"]
      }
    }
  }'

# Add Lambda target
aws events put-targets \
  --rule "CloudWatchAlarmStateChange" \
  --targets '[
    {
      "Id": "AlertHandler",
      "Arn": "arn:aws:lambda:ap-northeast-2:123456789012:function:alert-handler"
    }
  ]'

自動応答の設定

EventBridge イベントパターン

json

{
  "source": ["aws.cloudwatch"],
  "detail-type": ["CloudWatch Alarm State Change"],
  "detail": {
    "alarmName": [{
      "prefix": "EKS-"
    }],
    "state": {
      "value": ["ALARM"]
    },
    "previousState": {
      "value": ["OK"]
    },
    "configuration": {
      "metrics": [{
        "metricStat": {
          "metric": {
            "namespace": ["AWS/EKS", "ContainerInsights"]
          }
        }
      }]
    }
  }
}

自動復旧 Lambda の例

python

# auto_recovery.py
import boto3
import json

ec2 = boto3.client('ec2')
ecs = boto3.client('ecs')

def lambda_handler(event, context):
    alarm_name = event['detail']['alarmName']
    alarm_state = event['detail']['state']['value']

    print(f"Alarm: {alarm_name}, State: {alarm_state}")

    # Automatic response based on alarm name
    if 'EC2-HighCPU' in alarm_name:
        # Identify EC2 instance
        dimensions = event['detail']['configuration']['metrics'][0]['metricStat']['metric']['dimensions']
        instance_id = next(d['value'] for d in dimensions if d['name'] == 'InstanceId')

        # Reboot instance
        ec2.reboot_instances(InstanceIds=[instance_id])
        return {'action': 'reboot', 'instance': instance_id}

    elif 'ECS-ServiceUnhealthy' in alarm_name:
        # Restart ECS service
        dimensions = event['detail']['configuration']['metrics'][0]['metricStat']['metric']['dimensions']
        cluster = next(d['value'] for d in dimensions if d['name'] == 'ClusterName')
        service = next(d['value'] for d in dimensions if d['name'] == 'ServiceName')

        ecs.update_service(
            cluster=cluster,
            service=service,
            forceNewDeployment=True
        )
        return {'action': 'redeploy', 'service': service}

    return {'action': 'none'}

Container Insights アラート

EKS Container Insights メトリクス

Container Insights を有効にすると、EKS cluster のメトリクスを CloudWatch で確認できます。

bash

# Enable Container Insights
aws eks update-addon \
  --cluster-name my-cluster \
  --addon-name amazon-cloudwatch-observability \
  --addon-version v1.2.0-eksbuild.1

# Or install CloudWatch Agent
kubectl apply -f https://raw.githubusercontent.com/aws-samples/amazon-cloudwatch-container-insights/latest/k8s-deployment-manifest-templates/deployment-mode/daemonset/container-insights-monitoring/quickstart/cwagent-fluentd-quickstart.yaml

Container Insights アラートの例

bash

# Node CPU utilization alarm
aws cloudwatch put-metric-alarm \
  --alarm-name "EKS-Node-HighCPU" \
  --metric-name node_cpu_utilization \
  --namespace ContainerInsights \
  --dimensions Name=ClusterName,Value=my-cluster \
  --statistic Average \
  --period 300 \
  --threshold 80 \
  --comparison-operator GreaterThanThreshold \
  --evaluation-periods 2 \
  --alarm-actions arn:aws:sns:ap-northeast-2:123456789012:eks-alerts

# Pod memory utilization alarm
aws cloudwatch put-metric-alarm \
  --alarm-name "EKS-Pod-HighMemory" \
  --metric-name pod_memory_utilization \
  --namespace ContainerInsights \
  --dimensions Name=ClusterName,Value=my-cluster Name=Namespace,Value=production \
  --statistic Average \
  --period 300 \
  --threshold 85 \
  --comparison-operator GreaterThanThreshold \
  --evaluation-periods 2 \
  --alarm-actions arn:aws:sns:ap-northeast-2:123456789012:eks-alerts

# Pod restart alarm
aws cloudwatch put-metric-alarm \
  --alarm-name "EKS-Pod-Restarts" \
  --metric-name pod_number_of_container_restarts \
  --namespace ContainerInsights \
  --dimensions Name=ClusterName,Value=my-cluster Name=Namespace,Value=production \
  --statistic Sum \
  --period 300 \
  --threshold 3 \
  --comparison-operator GreaterThanThreshold \
  --evaluation-periods 1 \
  --alarm-actions arn:aws:sns:ap-northeast-2:123456789012:eks-alerts

主な Container Insights メトリクス

メトリクス	説明	ディメンション
`cluster_node_count`	Cluster の Node 数	ClusterName
`cluster_failed_node_count`	失敗した Node 数	ClusterName
`node_cpu_utilization`	Node CPU 使用率	ClusterName, NodeName
`node_memory_utilization`	Node メモリ使用率	ClusterName, NodeName
`node_filesystem_utilization`	Node ディスク使用率	ClusterName, NodeName
`pod_cpu_utilization`	Pod CPU 使用率	ClusterName, Namespace, PodName
`pod_memory_utilization`	Pod メモリ使用率	ClusterName, Namespace, PodName
`pod_number_of_container_restarts`	Container 再起動回数	ClusterName, Namespace, PodName
`service_number_of_running_pods`	Service ごとの実行中 Pod 数	ClusterName, Namespace, Service

CloudWatch Alarm アクション

EC2 アクション

bash

# EC2 instance recovery (on system status check failure)
aws cloudwatch put-metric-alarm \
  --alarm-name "EC2-SystemCheckFailed" \
  --metric-name StatusCheckFailed_System \
  --namespace AWS/EC2 \
  --dimensions Name=InstanceId,Value=i-1234567890abcdef0 \
  --statistic Maximum \
  --period 60 \
  --threshold 1 \
  --comparison-operator GreaterThanOrEqualToThreshold \
  --evaluation-periods 2 \
  --alarm-actions arn:aws:automate:ap-northeast-2:ec2:recover

# EC2 instance stop
aws cloudwatch put-metric-alarm \
  --alarm-name "EC2-LowUtilization-Stop" \
  --metric-name CPUUtilization \
  --namespace AWS/EC2 \
  --dimensions Name=InstanceId,Value=i-1234567890abcdef0 \
  --statistic Average \
  --period 3600 \
  --threshold 5 \
  --comparison-operator LessThanThreshold \
  --evaluation-periods 24 \
  --alarm-actions arn:aws:automate:ap-northeast-2:ec2:stop

Auto Scaling アクション

bash

# Link Auto Scaling policy
aws cloudwatch put-metric-alarm \
  --alarm-name "ASG-ScaleOut" \
  --metric-name CPUUtilization \
  --namespace AWS/EC2 \
  --dimensions Name=AutoScalingGroupName,Value=my-asg \
  --statistic Average \
  --period 300 \
  --threshold 70 \
  --comparison-operator GreaterThanThreshold \
  --evaluation-periods 2 \
  --alarm-actions arn:aws:autoscaling:ap-northeast-2:123456789012:scalingPolicy:xxx:autoScalingGroupName/my-asg:policyName/scale-out

aws cloudwatch put-metric-alarm \
  --alarm-name "ASG-ScaleIn" \
  --metric-name CPUUtilization \
  --namespace AWS/EC2 \
  --dimensions Name=AutoScalingGroupName,Value=my-asg \
  --statistic Average \
  --period 300 \
  --threshold 30 \
  --comparison-operator LessThanThreshold \
  --evaluation-periods 3 \
  --alarm-actions arn:aws:autoscaling:ap-northeast-2:123456789012:scalingPolicy:xxx:autoScalingGroupName/my-asg:policyName/scale-in

Systems Manager アクション

bash

# Execute SSM Automation
aws cloudwatch put-metric-alarm \
  --alarm-name "DiskFull-Cleanup" \
  --metric-name disk_used_percent \
  --namespace CWAgent \
  --dimensions Name=InstanceId,Value=i-1234567890abcdef0 Name=path,Value=/ \
  --statistic Average \
  --period 300 \
  --threshold 90 \
  --comparison-operator GreaterThanThreshold \
  --evaluation-periods 1 \
  --alarm-actions arn:aws:ssm:ap-northeast-2:123456789012:automation-definition/CleanupDisk:$DEFAULT

コスト最適化

コスト要因

項目	コスト
Standard Resolution アラーム（60 秒）	$0.10/アラーム/月
High Resolution アラーム（10 秒）	$0.30/アラーム/月
Anomaly Detection	$0.30/メトリクス/月
Composite Alarm	$0.50/アラーム/月

コスト最適化戦略

推奨設定

yaml

# Cost-effective alarm settings

# Critical: High Resolution (fast detection needed)
critical-alerts:
  period: 60  # 1 minute
  evaluation-periods: 2

# Warning: Standard Resolution
warning-alerts:
  period: 300  # 5 minutes
  evaluation-periods: 2

# Info: Standard Resolution (relaxed detection)
info-alerts:
  period: 900  # 15 minutes
  evaluation-periods: 3

アラームクリーンアップスクリプト

bash

#!/bin/bash
# Identify and clean up old alarms

# List alarms in INSUFFICIENT_DATA state for 90+ days
aws cloudwatch describe-alarms \
  --state-value INSUFFICIENT_DATA \
  --query 'MetricAlarms[?StateUpdatedTimestamp<=`2024-11-01`].AlarmName' \
  --output text

# Delete alarms
aws cloudwatch delete-alarms \
  --alarm-names "old-alarm-1" "old-alarm-2"

Prometheus Metrics 統合

Amazon Managed Prometheus (AMP) 統合

AMP メトリクスは CloudWatch アラートに使用できます。

bash

# Send AMP workspace metrics to CloudWatch
# (Periodic query via Lambda)

# Lambda function example

python

# amp_to_cloudwatch.py
import boto3
import requests
from aws_requests_auth.aws_auth import AWSRequestsAuth

def lambda_handler(event, context):
    # AMP workspace settings
    amp_endpoint = "https://aps-workspaces.ap-northeast-2.amazonaws.com/workspaces/ws-xxx/api/v1/query"
    region = "ap-northeast-2"

    # AWS authentication
    auth = AWSRequestsAuth(
        aws_access_key=boto3.Session().get_credentials().access_key,
        aws_secret_access_key=boto3.Session().get_credentials().secret_key,
        aws_token=boto3.Session().get_credentials().token,
        aws_host=f"aps-workspaces.{region}.amazonaws.com",
        aws_region=region,
        aws_service="aps"
    )

    # Execute Prometheus queries
    queries = [
        ("eks_node_cpu_usage", 'avg(rate(node_cpu_seconds_total{mode!="idle"}[5m])) * 100'),
        ("eks_pod_memory_usage", 'avg(container_memory_working_set_bytes) / avg(container_spec_memory_limit_bytes) * 100'),
    ]

    cloudwatch = boto3.client('cloudwatch')

    for metric_name, query in queries:
        response = requests.get(
            amp_endpoint,
            params={"query": query},
            auth=auth
        )

        result = response.json()
        if result['data']['result']:
            value = float(result['data']['result'][0]['value'][1])

            # Send metric to CloudWatch
            cloudwatch.put_metric_data(
                Namespace='AMP/EKS',
                MetricData=[{
                    'MetricName': metric_name,
                    'Value': value,
                    'Unit': 'Percent'
                }]
            )

    return {'status': 'success'}

Terraform の例

基本アラーム

hcl

# SNS Topic
resource "aws_sns_topic" "alerts" {
  name = "eks-alerts"
}

resource "aws_sns_topic_subscription" "email" {
  topic_arn = aws_sns_topic.alerts.arn
  protocol  = "email"
  endpoint  = "team@example.com"
}

# EC2 CPU alarm
resource "aws_cloudwatch_metric_alarm" "ec2_cpu" {
  alarm_name          = "ec2-high-cpu"
  comparison_operator = "GreaterThanThreshold"
  evaluation_periods  = 2
  metric_name         = "CPUUtilization"
  namespace           = "AWS/EC2"
  period              = 300
  statistic           = "Average"
  threshold           = 80
  alarm_description   = "EC2 CPU usage exceeds 80%"

  dimensions = {
    InstanceId = "i-1234567890abcdef0"
  }

  alarm_actions = [aws_sns_topic.alerts.arn]
  ok_actions    = [aws_sns_topic.alerts.arn]

  treat_missing_data = "notBreaching"
}

Metrics Math アラーム

hcl

resource "aws_cloudwatch_metric_alarm" "alb_error_rate" {
  alarm_name          = "alb-high-error-rate"
  comparison_operator = "GreaterThanThreshold"
  evaluation_periods  = 2
  threshold           = 5
  alarm_description   = "ALB error rate exceeds 5%"

  metric_query {
    id          = "errors"
    return_data = false

    metric {
      metric_name = "HTTPCode_Target_5XX_Count"
      namespace   = "AWS/ApplicationELB"
      period      = 300
      stat        = "Sum"

      dimensions = {
        LoadBalancer = "app/my-alb/1234567890"
      }
    }
  }

  metric_query {
    id          = "requests"
    return_data = false

    metric {
      metric_name = "RequestCount"
      namespace   = "AWS/ApplicationELB"
      period      = 300
      stat        = "Sum"

      dimensions = {
        LoadBalancer = "app/my-alb/1234567890"
      }
    }
  }

  metric_query {
    id          = "error_rate"
    expression  = "(errors / requests) * 100"
    label       = "Error Rate"
    return_data = true
  }

  alarm_actions = [aws_sns_topic.alerts.arn]
}

Composite Alarm

hcl

# Individual alarms
resource "aws_cloudwatch_metric_alarm" "cpu_alarm" {
  alarm_name          = "high-cpu"
  comparison_operator = "GreaterThanThreshold"
  evaluation_periods  = 2
  metric_name         = "CPUUtilization"
  namespace           = "AWS/EC2"
  period              = 300
  statistic           = "Average"
  threshold           = 80

  dimensions = {
    InstanceId = "i-1234567890abcdef0"
  }
}

resource "aws_cloudwatch_metric_alarm" "memory_alarm" {
  alarm_name          = "high-memory"
  comparison_operator = "GreaterThanThreshold"
  evaluation_periods  = 2
  metric_name         = "mem_used_percent"
  namespace           = "CWAgent"
  period              = 300
  statistic           = "Average"
  threshold           = 85

  dimensions = {
    InstanceId = "i-1234567890abcdef0"
  }
}

# Composite Alarm
resource "aws_cloudwatch_composite_alarm" "server_critical" {
  alarm_name        = "server-critical"
  alarm_description = "Server CPU and Memory are both high"

  alarm_rule = "ALARM(${aws_cloudwatch_metric_alarm.cpu_alarm.alarm_name}) AND ALARM(${aws_cloudwatch_metric_alarm.memory_alarm.alarm_name})"

  alarm_actions = [aws_sns_topic.alerts.arn]
  ok_actions    = [aws_sns_topic.alerts.arn]
}

EKS Container Insights アラーム

hcl

resource "aws_cloudwatch_metric_alarm" "eks_node_cpu" {
  alarm_name          = "eks-node-high-cpu"
  comparison_operator = "GreaterThanThreshold"
  evaluation_periods  = 2
  metric_name         = "node_cpu_utilization"
  namespace           = "ContainerInsights"
  period              = 300
  statistic           = "Average"
  threshold           = 80
  alarm_description   = "EKS Node CPU usage exceeds 80%"

  dimensions = {
    ClusterName = "my-eks-cluster"
  }

  alarm_actions = [aws_sns_topic.alerts.arn]
}

resource "aws_cloudwatch_metric_alarm" "eks_pod_restarts" {
  alarm_name          = "eks-pod-restarts"
  comparison_operator = "GreaterThanThreshold"
  evaluation_periods  = 1
  metric_name         = "pod_number_of_container_restarts"
  namespace           = "ContainerInsights"
  period              = 300
  statistic           = "Sum"
  threshold           = 3
  alarm_description   = "EKS Pod has restarted more than 3 times"

  dimensions = {
    ClusterName = "my-eks-cluster"
    Namespace   = "production"
  }

  alarm_actions = [aws_sns_topic.alerts.arn]
}

異常検知アラーム

hcl

resource "aws_cloudwatch_metric_alarm" "cpu_anomaly" {
  alarm_name          = "cpu-anomaly-detection"
  comparison_operator = "LessThanLowerOrGreaterThanUpperThreshold"
  evaluation_periods  = 2
  threshold_metric_id = "ad1"
  alarm_description   = "CPU usage is anomalous"

  metric_query {
    id          = "m1"
    return_data = true

    metric {
      metric_name = "CPUUtilization"
      namespace   = "AWS/EC2"
      period      = 300
      stat        = "Average"

      dimensions = {
        InstanceId = "i-1234567890abcdef0"
      }
    }
  }

  metric_query {
    id          = "ad1"
    expression  = "ANOMALY_DETECTION_BAND(m1, 2)"
    label       = "CPUUtilization (Expected)"
    return_data = true
  }

  alarm_actions = [aws_sns_topic.alerts.arn]
}

クイズ

CloudWatch Alarms クイズで理解度を確認しましょう。

エンドツーエンドのオブザーバビリティラボ

ArgoCD の詳細クイズ

Istio の詳細クイズ

Linkerd の詳細クイズ

Cilium サービスメッシュの詳細クイズ

Cilium 詳説

Calico 詳説

メトリクス

ログ記録

トレーシング

アラート通知

可観測性の演習

ArgoCD

トラフィック管理

セキュリティ

可観測性

レジリエンス

高度な設定

比較ガイド

メトリクス

ログ記録

トレーシング

アラート通知

CloudWatch Alarms ​

目次 ​

CloudWatch Alarms の概要 ​

主な機能 ​

CloudWatch Alarms と Prometheus Alertmanager の比較 ​

アーキテクチャ ​

CloudWatch Alarms の動作フロー ​

アラームの状態 ​

Metric Alarms ​

基本的なアラーム作成（Console/CLI） ​

AWS CLI ​

アラーム設定の構成要素 ​

比較演算子 ​

Metrics Math を使用するアラーム ​

Metrics Math 関数 ​

Composite Alarms ​

Composite Alarm の概念 ​

Composite Alarms の作成 ​

アラームルールの構文 ​

アラート抑制パターン ​

異常検知 ​

異常検知の概要 ​

異常検知アラームの作成 ​

異常検知の設定 ​

モデル学習期間の調整 ​

SNS 統合 ​

SNS Topic の作成 ​

SNS メッセージフィルタリング ​

SNS から Slack への統合（Lambda） ​

EventBridge 統合 ​

EventBridge ルールの作成 ​

自動応答の設定 ​

EventBridge イベントパターン ​

自動復旧 Lambda の例 ​

Container Insights アラート ​

EKS Container Insights メトリクス ​

Container Insights アラートの例 ​

主な Container Insights メトリクス ​

CloudWatch Alarm アクション ​

EC2 アクション ​

Auto Scaling アクション ​

Systems Manager アクション ​

コスト最適化 ​

コスト要因 ​

コスト最適化戦略 ​

推奨設定 ​

アラームクリーンアップスクリプト ​

Prometheus Metrics 統合 ​

Amazon Managed Prometheus (AMP) 統合 ​

Terraform の例 ​

基本アラーム ​

Metrics Math アラーム ​

Composite Alarm ​

EKS Container Insights アラーム ​

異常検知アラーム ​

クイズ ​

CloudWatch Alarms

目次

CloudWatch Alarms の概要

主な機能

CloudWatch Alarms と Prometheus Alertmanager の比較

アーキテクチャ

CloudWatch Alarms の動作フロー

アラームの状態

Metric Alarms

基本的なアラーム作成（Console/CLI）

AWS CLI

アラーム設定の構成要素

比較演算子

Metrics Math を使用するアラーム

Metrics Math 関数

Composite Alarms

Composite Alarm の概念

Composite Alarms の作成

アラームルールの構文

アラート抑制パターン

異常検知

異常検知の概要

異常検知アラームの作成

異常検知の設定

モデル学習期間の調整

SNS 統合

SNS Topic の作成

SNS メッセージフィルタリング

SNS から Slack への統合（Lambda）

EventBridge 統合

EventBridge ルールの作成

自動応答の設定

EventBridge イベントパターン

自動復旧 Lambda の例

Container Insights アラート

EKS Container Insights メトリクス

Container Insights アラートの例

主な Container Insights メトリクス

CloudWatch Alarm アクション

EC2 アクション

Auto Scaling アクション

Systems Manager アクション

コスト最適化

コスト要因

コスト最適化戦略

推奨設定

アラームクリーンアップスクリプト

Prometheus Metrics 統合

Amazon Managed Prometheus (AMP) 統合

Terraform の例

基本アラーム

Metrics Math アラーム

Composite Alarm

EKS Container Insights アラーム

異常検知アラーム

クイズ