codezhong

K8s Scraping Hpa Best Practices

2026/05/08 共 5271 字，约 16 分钟

K8S监控采集与HPA配置：弹性伸缩实战指南

情境与背景

监控数据采集频率和HPA（Horizontal Pod Autoscaler）配置是Kubernetes弹性伸缩的核心要素。作为高级DevOps/SRE工程师，必须掌握合理的采集策略和HPA配置方法。本文从DevOps/SRE视角，详细讲解监控采集频率设置和HPA配置的最佳实践。

一、监控数据采集频率

1.1 采集频率配置

配置示例：

# Prometheus采集配置
scrape_configs:
  - job_name: "kubernetes-nodes"
    scrape_interval: "15s"
    scrape_timeout: "10s"
    static_configs:
      - targets: ["node-exporter:9100"]
  
  - job_name: "kubernetes-pods"
    scrape_interval: "30s"
    scrape_timeout: "20s"
    kubernetes_sd_configs:
      - role: pod
  
  - job_name: "custom-metrics"
    scrape_interval: "10s"
    scrape_timeout: "5s"
    static_configs:
      - targets: ["api-service:8080"]

1.2 采集频率策略

策略配置：

# 采集频率策略
scrape_strategy:
  critical:
    interval: "10s"
    timeout: "5s"
    description: "关键业务指标"
  
  important:
    interval: "15s"
    timeout: "10s"
    description: "核心服务指标"
  
  normal:
    interval: "30s"
    timeout: "20s"
    description: "常规服务指标"
  
  low:
    interval: "60s"
    timeout: "30s"
    description: "非关键指标"

1.3 指标分类与频率

指标分类：

指标类型采集频率说明
节点指标15-30sCPU、内存、磁盘、网络
K8S状态指标30-60sPod状态、Deployment状态
应用业务指标10-15sQPS、延迟、错误率
日志指标60s日志计数、错误数
自定义指标10-30s根据业务重要性

二、HPA配置详解

2.1 CPU HPA配置

配置示例：

# CPU HPA配置
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: cpu-hpa
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: my-app
  minReplicas: 2
  maxReplicas: 10
  metrics:
    - type: Resource
      resource:
        name: cpu
        target:
          type: Utilization
          averageUtilization: 70

2.2 内存HPA配置

配置示例：

# 内存HPA配置
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: memory-hpa
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: my-app
  minReplicas: 2
  maxReplicas: 10
  metrics:
    - type: Resource
      resource:
        name: memory
        target:
          type: Utilization
          averageUtilization: 75

2.3 混合指标HPA配置

配置示例：

# 混合指标HPA配置
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: mixed-hpa
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: my-app
  minReplicas: 2
  maxReplicas: 10
  metrics:
    - type: Resource
      resource:
        name: cpu
        target:
          type: Utilization
          averageUtilization: 70
    - type: Resource
      resource:
        name: memory
        target:
          type: Utilization
          averageUtilization: 75

2.4 自定义指标HPA配置

配置示例：

# 自定义指标HPA配置
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: custom-hpa
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: my-app
  minReplicas: 2
  maxReplicas: 10
  metrics:
    - type: Pods
      pods:
        metric:
          name: http_requests_per_second
        target:
          type: AverageValue
          averageValue: "100"

三、HPA参数调优

3.1 扩缩容策略

策略配置：

# 扩缩容策略
behavior:
  scaleUp:
    stabilizationWindowSeconds: 30
    policies:
      - type: Percent
        value: 100
        periodSeconds: 60
      - type: Pods
        value: 4
        periodSeconds: 60
    selectPolicy: Max
  
  scaleDown:
    stabilizationWindowSeconds: 600
    policies:
      - type: Percent
        value: 10
        periodSeconds: 60
      - type: Pods
        value: 1
        periodSeconds: 60
    selectPolicy: Min

3.2 参数说明

参数详解：

参数说明建议值
minReplicas最小副本数2+（保证高可用）
maxReplicas最大副本数根据业务峰值
stabilizationWindowSeconds稳定窗口扩容30-60s，缩容300-600s
scaleUp policies扩容策略百分比或固定数量
scaleDown policies缩容策略保守策略

四、监控与HPA集成

4.1 指标采集与HPA

集成架构：

# 监控与HPA集成
integration:
  prometheus:
    service: "prometheus-service"
    port: 9090
  
  metrics_server:
    enabled: true
    resources:
      requests:
        cpu: "100m"
        memory: "200Mi"
  
  custom_metrics:
    api_server:
      enabled: true
      service: "custom-metrics-apiserver"

4.2 Metrics Server配置

配置示例：

# Metrics Server配置
apiVersion: v1
kind: ServiceAccount
metadata:
  name: metrics-server
  namespace: kube-system
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: metrics-server
  namespace: kube-system
spec:
  replicas: 1
  selector:
    matchLabels:
      k8s-app: metrics-server
  template:
    metadata:
      labels:
        k8s-app: metrics-server
    spec:
      serviceAccountName: metrics-server
      containers:
        - name: metrics-server
          image: k8s.gcr.io/metrics-server/metrics-server:v0.6.4
          args:
            - "--cert-dir=/tmp"
            - "--secure-port=4443"
            - "--kubelet-preferred-address-types=InternalIP"

五、最佳实践

5.1 采集频率优化

优化策略：

# 采集频率优化策略
optimization:
  reduce_scrape_load:
    - "合并相似的采集任务"
    - "使用标签选择器过滤"
    - "设置合理的采集超时"
  
  prioritize_critical:
    - "关键指标高频采集"
    - "非关键指标低频采集"
    - "按需采集（基于标签）"

5.2 HPA配置最佳实践

配置建议：

# HPA最佳实践
hpa_best_practices:
  min_replicas:
    recommendation: "至少2个副本"
    reason: "保证高可用"
  
  max_replicas:
    recommendation: "根据业务峰值计算"
    formula: "峰值QPS / 单Pod处理能力"
  
  target_utilization:
    cpu: "60-80%"
    memory: "70-85%"
  
  scale_down_delay:
    recommendation: "至少5分钟"
    reason: "避免频繁扩缩容"

5.3 避免的坑

常见问题：

问题原因解决方案
采集频率过高资源消耗大根据指标重要性分级
HPA震荡阈值设置不合理调整稳定窗口和策略
扩缩容不及时指标延迟优化采集频率
资源不足集群资源有限配置Cluster Autoscaler

六、实战案例分析

6.1 案例1：高频采集优化

场景描述：

  • 监控资源消耗过大
  • 需要优化采集频率

优化方案：

# 优化后的采集配置
scrape_configs:
  - job_name: "critical-services"
    scrape_interval: "10s"
    static_configs:
      - targets: ["api-gateway:8080"]
  
  - job_name: "normal-services"
    scrape_interval: "30s"
    kubernetes_sd_configs:
      - role: pod
        namespaces:
          names: ["default"]
  
  - job_name: "monitoring-services"
    scrape_interval: "60s"
    static_configs:
      - targets: ["node-exporter:9100"]

6.2 案例2：HPA配置优化

场景描述：

  • HPA频繁扩缩容
  • 需要优化配置

优化方案：

# 优化后的HPA配置
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: optimized-hpa
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: my-app
  minReplicas: 3
  maxReplicas: 15
  metrics:
    - type: Resource
      resource:
        name: cpu
        target:
          type: Utilization
          averageUtilization: 75
  behavior:
    scaleUp:
      stabilizationWindowSeconds: 60
      policies:
        - type: Percent
          value: 50
          periodSeconds: 60
    scaleDown:
      stabilizationWindowSeconds: 900
      policies:
        - type: Percent
          value: 5
          periodSeconds: 300

七、面试1分钟精简版（直接背）

完整版：

监控数据采集频率根据指标类型有所不同，Prometheus默认是15秒，关键业务指标可以配置为10-15秒，非关键指标可以设为30-60秒。HPA配置方面，我们配置基于CPU和内存的自动扩缩容，目标阈值分别设为70%和75%，最小副本数设置为2保证高可用，最大副本数根据业务峰值设置。同时支持基于自定义指标的HPA，如QPS、请求延迟等。

30秒超短版：

采集频率15-60秒，关键指标高频。HPA配置CPU/内存指标，最小2副本，阈值70-75%。

八、总结

8.1 核心要点

  1. 采集频率：根据指标重要性分级设置（10s-60s）
  2. HPA配置：CPU目标60-80%，内存目标70-85%
  3. 最小副本：至少2个保证高可用
  4. 最大副本：根据业务峰值计算
  5. 扩缩容策略：扩容激进，缩容保守

8.2 配置原则

原则说明
分级采集关键指标高频，非关键指标低频
高可用优先最小副本数>=2
避免震荡设置合理的稳定窗口
弹性伸缩结合Cluster Autoscaler

8.3 记忆口诀

采集频率分等级，15到60秒合适，
HPA配置看指标，CPU内存双保障，
最小副本要保证，最大副本有上限，
扩容激进缩容稳，弹性伸缩保稳定。

参考链接：SRE运维面试题全解析：从理论到实践（第二部分）

文档信息

Search

    Table of Contents