k8s中安装metrics，实现hpa

码上上班2026-03-25 11:31

1. HPA 是什么？

全称：Horizontal Pod Autoscaler 中文：Pod 水平自动扩缩容

一句话解释：根据 CPU / 内存使用率，自动增加 / 减少 Pod 数量！

2. 为什么要用 HPA？

  • 流量高 → 自动多开 Pod
  • 流量低 → 自动关掉多余 Pod
  • 不用人工手动扩缩容
  • 防止服务崩溃、节省资源

3. HPA 靠什么工作？

必须依赖 metrics-server→ 也就是你刚才一直在装的组件！

HPA 工作流程：

  1. metrics-server 采集 Pod CPU / 内存
  2. HPA 读取指标
  3. 自动判断是否扩容 / 缩容

=====================================

操作步骤

1.metrics-server文件(用户采集信息)

复制代码 
vim components.yaml

apiVersion: v1
kind: ServiceAccount
metadata:
  labels:
    k8s-app: metrics-server-2
  name: metrics-server-2
  namespace: kube-system
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  labels:
    k8s-app: metrics-server-2
    rbac.authorization.k8s.io/aggregate-to-admin: "true"
    rbac.authorization.k8s.io/aggregate-to-edit: "true"
    rbac.authorization.k8s.io/aggregate-to-view: "true"
  name: system:aggregated-metrics-reader
rules:
- apiGroups:
  - metrics.k8s.io
  resources:
  - pods
  - nodes
  verbs:
  - get
  - list
  - watch
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  labels:
    k8s-app: metrics-server-2
  name: system:metrics-server-2
rules:
- apiGroups:
  - ""
  resources:
  - nodes/metrics
  verbs:
  - get
- apiGroups:
  - ""
  resources:
  - pods
  - nodes
  verbs:
  - get
  - list
  - watch
---
apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
  labels:
    k8s-app: metrics-server-2
  name: metrics-server-2-auth-reader
  namespace: kube-system
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: Role
  name: extension-apiserver-authentication-reader
subjects:
- kind: ServiceAccount
  name: metrics-server-2
  namespace: kube-system
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  labels:
    k8s-app: metrics-server-2
  name: metrics-server-2:system:auth-delegator
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: ClusterRole
  name: system:auth-delegator
subjects:
- kind: ServiceAccount
  name: metrics-server-2
  namespace: kube-system
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  labels:
    k8s-app: metrics-server-2
  name: system:metrics-server-2
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: ClusterRole
  name: system:metrics-server-2
subjects:
- kind: ServiceAccount
  name: metrics-server-2
  namespace: kube-system
---
apiVersion: v1
kind: Service
metadata:
  labels:
    k8s-app: metrics-server-2
  name: metrics-server-2
  namespace: kube-system
spec:
  ports:
  - name: https
    port: 443
    protocol: TCP
    targetPort: https
  selector:
    k8s-app: metrics-server-2
---
apiVersion: apps/v1
kind: Deployment
metadata:
  labels:
    k8s-app: metrics-server-2
  name: metrics-server-2
  namespace: kube-system
spec:
  replicas: 2
  selector:
    matchLabels:
      k8s-app: metrics-server-2
  strategy:
    rollingUpdate:
      maxUnavailable: 1
  template:
    metadata:
      labels:
        k8s-app: metrics-server-2
    spec:
      affinity:
        podAntiAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
          - labelSelector:
              matchLabels:
                k8s-app: metrics-server-2
            namespaces:
            - kube-system
            topologyKey: kubernetes.io/hostname
      containers:
      - args:
        - --cert-dir=/tmp
        - --secure-port=10250
        - --kubelet-preferred-address-types=InternalIP,Hostname,InternalDNS,ExternalDNS,ExternalIP
        - --kubelet-use-node-status-port
        - --metric-resolution=15s
        - --kubelet-insecure-tls  #跳过证书，能连节点 自建 K8s 集群的节点证书不被系统信任
        - --kubelet-preferred-address-types=InternalIP,Hostname,InternalDNS,ExternalDNS,ExternalIP   #优先用内网 IP 连接，保证稳定
        image: swr.cn-north-4.myhuaweicloud.com/ddn-k8s/registry.k8s.io/metrics-server/metrics-server:v0.7.2  #换镜像 = 国内能下载
        imagePullPolicy: IfNotPresent
        livenessProbe:
          failureThreshold: 3
          httpGet:
            path: /livez
            port: https
            scheme: HTTPS
          periodSeconds: 10
        name: metrics-server-2
        ports:
        - containerPort: 10250
          name: https
          protocol: TCP
        readinessProbe:
          failureThreshold: 3
          httpGet:
            path: /readyz
            port: https
            scheme: HTTPS
          initialDelaySeconds: 20
          periodSeconds: 10
        resources:
          requests:
            cpu: 100m
            memory: 200Mi
        securityContext:
          allowPrivilegeEscalation: false
          capabilities:
            drop:
            - ALL
          readOnlyRootFilesystem: true
          runAsNonRoot: true
          runAsUser: 1000
          seccompProfile:
            type: RuntimeDefault
        volumeMounts:
        - mountPath: /tmp
          name: tmp-dir
      nodeSelector:
        kubernetes.io/os: linux
      priorityClassName: system-cluster-critical
      serviceAccountName: metrics-server-2
      volumes:
      - emptyDir: {}
        name: tmp-dir
---
apiVersion: policy/v1
kind: PodDisruptionBudget
metadata:
  labels:
    k8s-app: metrics-server-2
  name: metrics-server-2
  namespace: kube-system
spec:
  minAvailable: 1
  selector:
    matchLabels:
      k8s-app: metrics-server-2
---
apiVersion: apiregistration.k8s.io/v1
kind: APIService
metadata:
  labels:
    k8s-app: metrics-server-2
  name: v1beta1.metrics.k8s.io
spec:
  group: metrics.k8s.io
  groupPriorityMinimum: 100
  insecureSkipTLSVerify: true
  service:
    name: metrics-server-2
    namespace: kube-system
  version: v1beta1
  versionPriority: 100

2.安装：kubectl apply -f components.yaml

3.查看是否安装成功：kubectl top nodes --use-protocol-buffers

4.HAP监控Cpu使用率

复制代码 
vim k8s-hpa-deploy-svc.yaml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: nginx-deploy
spec:
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
        - name: nginx
          image: nginx:1.17.1
          ports:
            - containerPort: 80
          resources: # 资源限制
            requests:
              cpu: "100m" # 100m 表示100 milli cpu，即 0.1 个CPU
---
apiVersion: v1
kind: Service
metadata:
  name: nginx-svc
spec:
  selector:
    app: nginx
  type: NodePort
  ports:
    - port: 80 # svc 的访问端口
      name: nginx
      targetPort: 80 # Pod 的访问端口
      protocol: TCP
      nodePort: 30010 # 在机器上开端口，浏览器访问

kubectl apply -f k8s-hpa-deploy-svc.yaml

创建HPA

复制代码 
vim k8s-hpa.yaml

apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: hpa-cpu
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: nginx-deploy
  minReplicas: 1
  maxReplicas: 10
  metrics:
  - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 10  # CPU 50%

测试

复制代码 
kubectl run -i --tty load-generator --rm --image=busybox --restart=Never -- /bin/sh -c "while sleep 0.01; do wget -q -O- http://192.168.118.100:30010; done"

5.HAP监控内存使用率

(记得先删除多余的hpa : kubectl delete hpa <hpa名字>)

创建HPA

复制代码 
vim k8s-neicun.yaml

apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: hpa-memory
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: nginx-deploy
  minReplicas: 1
  maxReplicas: 10
  metrics:
  - type: Resource
    resource:
      name: memory
      target:
        type: Utilization
        averageUtilization: 50  # 内存 50%

kubectl apply -f  k8s-neicun.yaml

测试

复制代码 
kubectl run -i --tty load-generator --rm --image=busybox --restart=Never -- /bin/sh -c "while sleep 0.01; do wget -q -O- http://192.168.118.100:30010; done"

6.同时监控 CPU + 内存（最常用）

复制代码 
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: hpa-cpu-mem
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: nginx-deploy
  minReplicas: 1
  maxReplicas: 10
  metrics:
  - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 50
  - type: Resource
    resource:
      name: memory
      target:
        type: Utilization
        averageUtilization: 50

7.设置扩缩容速度参数(控制扩缩容等待时间)

(记得先删除多余的hpa : kubectl delete hpa <hpa名字>)

K8s 为了避免频繁扩缩容导致服务抖动，给 HPA 加了两个关键冷却时间：

  • 扩容冷却 ：默认 3 分钟 → 刚扩容后，3 分钟内不会再次扩容

  • 缩容冷却 ：默认 5 分钟 → 负载降下来后，必须等 5 分钟才会开始缩容

    behavior:
    scaleUp:
    stabilizationWindowSeconds: 60 # 扩容等待时间（避免频繁扩容）
    scaleDown:
    stabilizationWindowSeconds: 60 # 缩容等待时间（改成 60 秒测试很快看到缩容）

完整yaml

复制代码 
vim k8s-hpa-windows.yaml

apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: test-hpa
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: nginx-deploy
  minReplicas: 1
  maxReplicas: 10
  metrics:
    - type: Resource
      resource:
        name: cpu
        target:
          type: Utilization
          averageUtilization: 10  # 阈值低，测试容易触发扩容
  behavior:
    scaleUp:
      stabilizationWindowSeconds: 10  # 扩容等待时间（避免频繁扩容）
    scaleDown:
      stabilizationWindowSeconds: 10  # 缩容等待时间（改成 10 秒测试很快看到缩容）

kubectl apply -f k8s-hpa-windows.yaml

效果很快10秒就出效果

压力测另一种方式ab

复制代码 
kubectl run -it --rm ab --image=ikane/httpd-tools -- ab -c 10 -n 100000 http://192.168.118.100:30010/

===========================
yum install -y httpd-tools
#安装

ab -c 10 -n 100000 http://192.168.118.100:30010/
#-c 10 一次性并发访问10条
#-n 100000 总共发送100000条
相关推荐
Eine .5 小时前
Docker容器技术
运维·docker·容器
code_pgf5 小时前
Orin NX 16GB 的 package 安装命令清单 + Docker/工作区目录结构 + bringup 顺序
运维·docker·容器·ros
biuyyyxxx6 小时前
Power Query功能区 - 主页
笔记·学习·excel
sky wide7 小时前
[特殊字符] Docker 安装指南 (CentOS)
docker·容器·centos
CDA数据分析师干货分享7 小时前
【经验贴】机械工程本科，CDA数据分析师学习及转行经验
数据挖掘·数据分析·excel·cda证书·cda数据分析师
何妨呀~8 小时前
Docker部署与配置全攻略
运维·docker·容器
河码匠9 小时前
Kubernetes YAML 详解之网络服务二（ Ingress、IngressClasses）
云原生·容器·kubernetes
blackorbird9 小时前
一个来自法国的基于K8s的规模化扫描集群
云原生·容器·kubernetes
小邓睡不饱耶10 小时前
东方财富股吧话题爬虫实现：从接口请求到Excel数据落地
爬虫·excel
风向决定发型丶10 小时前
浅谈K8S的Label和Annotation
云原生·容器·kubernetes
热门推荐
012026年3月AI领域大事件：DeepSeek引领开源风暴02GitHub 镜像站点03围棋-html版本04Qwen3.5 开源全解析：从 0.8B 到 397B，代际升级 + 全场景选型指南05小黑课堂计算机二级WPSoffice题库软件下载安装教程（2026年3月最新版）06班级宠物园部署指南07【计算机一级WPSoffice】小黑课堂题库软件下载安装教程（2026年3月最新版）08UV安装并设置国内源09OpenClaw 使用和管理 MCP 完全指南10让 Trae IDE 智能体 “读懂”文档 Excel+PDF+DOCX ：mcp-documents-reader 工具使用指南