K8s + SpringBoot实现零宕机发布

Posted 程序员大咖

tags:

篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了K8s + SpringBoot实现零宕机发布相关的知识,希望对你有一定的参考价值。

👇👇关注后回复 “进群” ,拉你进程序员交流群👇👇

来源:blog.csdn.net/qq_14999375/article/details/123309636

# 前言

K8s + SpringBoot实现零宕机发布:健康检查+滚动更新+优雅停机+弹性伸缩+Prometheus监控+配置分离(镜像复用)

# 配置

健康检查

  • 健康检查类型:就绪探针(readiness)+ 存活探针(liveness)

  • 探针类型:exec(进入容器执行脚本)、tcpSocket(探测端口)、httpGet(调用接口)


业务层面

项目依赖 pom.xml

<dependency>
    <groupId>org.springframework.boot</groupId>
    <artifactId>spring-boot-starter-actuator</artifactId>
</dependency>

定义访问端口、路径及权限 application.yaml

management:
  server:
    port: 50000                         # 启用独立运维端口
  endpoint:                             # 开启health端点
    health:
      probes:
        enabled: true
  endpoints:
    web:
      exposure:
        base-path: /actuator            # 指定上下文路径,启用相应端点
        include: health

将暴露/actuator/health/readiness和/actuator/health/liveness两个接口,访问方式如下:

http://127.0.0.1:50000/actuator/health/readiness
http://127.0.0.1:50000/actuator/health/liveness
运维层面

k8s部署模版deployment.yaml

apiVersion: apps/v1
kind: Deployment
spec:
  template:
    spec:
      containers:
      - name: APP_NAME
        image: IMAGE_URL
        imagePullPolicy: Always
        ports:
        - containerPort: APP_PORT
        - name: management-port
          containerPort: 50000         # 应用管理端口
        readinessProbe:                # 就绪探针
          httpGet:
            path: /actuator/health/readiness
            port: management-port
          initialDelaySeconds: 30      # 延迟加载时间
          periodSeconds: 10            # 重试时间间隔
          timeoutSeconds: 1            # 超时时间设置
          successThreshold: 1          # 健康阈值
          failureThreshold: 6          # 不健康阈值
        livenessProbe:                 # 存活探针
          httpGet:
            path: /actuator/health/liveness
            port: management-port
          initialDelaySeconds: 30      # 延迟加载时间
          periodSeconds: 10            # 重试时间间隔
          timeoutSeconds: 1            # 超时时间设置
          successThreshold: 1          # 健康阈值
          failureThreshold: 6          # 不健康阈值

滚动更新

k8s资源调度之滚动更新策略,若要实现零宕机发布,需支持健康检查

apiVersion: apps/v1
kind: Deployment
metadata:
  name: APP_NAME
  labels:
    app: APP_NAME
spec:
  selector:
    matchLabels:
      app: APP_NAME
  replicas: REPLICAS    # Pod副本数
  strategy:
    type: RollingUpdate    # 滚动更新策略
    rollingUpdate:
      maxSurge: 1                   # 升级过程中最多可以比原先设置的副本数多出的数量
      maxUnavailable: 1             # 升级过程中最多有多少个POD处于无法提供服务的状态
优雅停机

在K8s中,当我们实现滚动升级之前,务必要实现应用级别的优雅停机。否则滚动升级时,还是会影响到业务。使应用关闭线程、释放连接资源后再停止服务


业务层面

项目依赖 pom.xml

<dependency>
    <groupId>org.springframework.boot</groupId>
    <artifactId>spring-boot-starter-actuator</artifactId>
</dependency>

定义访问端口、路径及权限 application.yaml

spring:
  application:
    name: <xxx>
  profiles:
    active: @profileActive@
  lifecycle:
    timeout-per-shutdown-phase: 30s     # 停机过程超时时长设置30s,超过30s,直接停机


server:
  port: 8080
  shutdown: graceful                    # 默认为IMMEDIATE,表示立即关机;GRACEFUL表示优雅关机


management:
  server:
    port: 50000                         # 启用独立运维端口
  endpoint:                             # 开启shutdown和health端点
    shutdown:
      enabled: true
    health:
      probes:
        enabled: true
  endpoints:
    web:
      exposure:
        base-path: /actuator            # 指定上下文路径,启用相应端点
        include: health,shutdown

将暴露/actuator/shutdown接口,调用方式如下:

curl -X POST 127.0.0.1:50000/actuator/shutdown
运维层面

确保dockerfile模版集成curl工具,否则无法使用curl命令

FROM openjdk:8-jdk-alpine
#构建参数
ARG JAR_FILE
ARG WORK_PATH="/app"
ARG EXPOSE_PORT=8080


#环境变量
ENV JAVA_OPTS=""\\
    JAR_FILE=$JAR_FILE


#设置时区
RUN ln -sf /usr/share/zoneinfo/Asia/Shanghai /etc/localtime && echo 'Asia/Shanghai' >/etc/timezone
RUN sed -i 's/dl-cdn.alpinelinux.org/mirrors.ustc.edu.cn/g' /etc/apk/repositories  \\
    && apk add --no-cache curl
#将maven目录的jar包拷贝到docker中,并命名为for_docker.jar
COPY target/$JAR_FILE $WORK_PATH/




#设置工作目录
WORKDIR $WORK_PATH




# 指定于外界交互的端口
EXPOSE $EXPOSE_PORT
# 配置容器,使其可执行化
ENTRYPOINT exec java $JAVA_OPTS -jar $JAR_FILE

k8s部署模版deployment.yaml

注:经验证,java项目可省略结束回调钩子的配置

此外,若需使用回调钩子,需保证镜像中包含curl工具,且需注意应用管理端口(50000)不能暴露到公网

apiVersion: apps/v1
kind: Deployment
spec:
  template:
    spec:
      containers:
      - name: APP_NAME
        image: IMAGE_URL
        imagePullPolicy: Always
        ports:
        - containerPort: APP_PORT
        - containerPort: 50000
        lifecycle:
          preStop:       # 结束回调钩子
            exec:
              command: ["curl", "-XPOST", "127.0.0.1:50000/actuator/shutdown"]

弹性伸缩

为pod设置资源限制后,创建HPA

apiVersion: apps/v1
kind: Deployment
metadata:
  name: APP_NAME
  labels:
    app: APP_NAME
spec:
  template:
    spec:
      containers:
      - name: APP_NAME
        image: IMAGE_URL
        imagePullPolicy: Always
        resources:                     # 容器资源管理
          limits:                      # 资源限制(监控使用情况)
            cpu: 0.5
            memory: 1Gi
          requests:                    # 最小可用资源(灵活调度)
            cpu: 0.15
            memory: 300Mi
---
kind: HorizontalPodAutoscaler            # 弹性伸缩控制器
apiVersion: autoscaling/v2beta2
metadata:
  name: APP_NAME
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: APP_NAME
  minReplicas: REPLICAS                # 缩放范围
  maxReplicas: 6
  metrics:
    - type: Resource
      resource:
        name: cpu                        # 指定资源指标
        target:
          type: Utilization
          averageUtilization: 50

Prometheus集成


业务层面

项目依赖 pom.xml

<!-- 引入Spring boot的监控机制-->
<dependency>
    <groupId>org.springframework.boot</groupId>
    <artifactId>spring-boot-starter-actuator</artifactId>
</dependency>
<dependency>
    <groupId>io.micrometer</groupId>
    <artifactId>micrometer-registry-prometheus</artifactId>
</dependency>

定义访问端口、路径及权限 application.yaml

management:
  server:
    port: 50000                         # 启用独立运维端口
  metrics:
    tags:
      application: $spring.application.name
  endpoints:
    web:
      exposure:
        base-path: /actuator            # 指定上下文路径,启用相应端点
        include: metrics,prometheus

将暴露/actuator/metric和/actuator/prometheus接口,访问方式如下:

http://127.0.0.1:50000/actuator/metric
http://127.0.0.1:50000/actuator/prometheus
运维层面

deployment.yaml

apiVersion: apps/v1
kind: Deployment
spec:
  template:
    metadata:
      annotations:
        prometheus:io/port: "50000"
        prometheus.io/path: /actuator/prometheus  # 在流水线中赋值
        prometheus.io/scrape: "true"              # 基于pod的服务发现

配置分离

方案:通过configmap挂载外部配置文件,并指定激活环境运行

作用:配置分离,避免敏感信息泄露;镜像复用,提高交付效率

通过文件生成configmap

# 通过dry-run的方式生成yaml文件
kubectl create cm -n <namespace> <APP_NAME> --from-file=application-test.yaml --dry-run=1 -oyaml > configmap.yaml


# 更新
kubectl apply -f configmap.yaml

挂载configmap并指定激活环境

apiVersion: apps/v1
kind: Deployment
metadata:
  name: APP_NAME
  labels:
    app: APP_NAME
spec:
  template:
    spec:
      containers:
      - name: APP_NAME
        image: IMAGE_URL
        imagePullPolicy: Always
        env:
          - name: SPRING_PROFILES_ACTIVE   # 指定激活环境
            value: test
        volumeMounts:                      # 挂载configmap
        - name: conf
          mountPath: "/app/config"         # 与Dockerfile中工作目录一致
          readOnly: true
      volumes:
      - name: conf
        configMap:
          name: APP_NAME

# 汇总配置

业务层面

项目依赖 pom.xml

<!-- 引入Spring boot的监控机制-->
<dependency>
    <groupId>org.springframework.boot</groupId>
    <artifactId>spring-boot-starter-actuator</artifactId>
</dependency>
<dependency>
    <groupId>io.micrometer</groupId>
    <artifactId>micrometer-registry-prometheus</artifactId>
</dependency>

定义访问端口、路径及权限 application.yaml

spring:
  application:
    name: project-sample
  profiles:
    active: @profileActive@
  lifecycle:
    timeout-per-shutdown-phase: 30s     # 停机过程超时时长设置30s,超过30s,直接停机


server:
  port: 8080
  shutdown: graceful                    # 默认为IMMEDIATE,表示立即关机;GRACEFUL表示优雅关机


management:
  server:
    port: 50000                         # 启用独立运维端口
  metrics:
    tags:
      application: $spring.application.name
  endpoint:                             # 开启shutdown和health端点
    shutdown:
      enabled: true
    health:
      probes:
        enabled: true
  endpoints:
    web:
      exposure:
        base-path: /actuator            # 指定上下文路径,启用相应端点
        include: health,shutdown,metrics,prometheus

运维层面

确保dockerfile模版集成curl工具,否则无法使用curl命令

FROM openjdk:8-jdk-alpine
#构建参数
ARG JAR_FILE
ARG WORK_PATH="/app"
ARG EXPOSE_PORT=8080


#环境变量
ENV JAVA_OPTS=""\\
    JAR_FILE=$JAR_FILE


#设置时区
RUN ln -sf /usr/share/zoneinfo/Asia/Shanghai /etc/localtime && echo 'Asia/Shanghai' >/etc/timezone
RUN sed -i 's/dl-cdn.alpinelinux.org/mirrors.ustc.edu.cn/g' /etc/apk/repositories  \\
    && apk add --no-cache curl
#将maven目录的jar包拷贝到docker中,并命名为for_docker.jar
COPY target/$JAR_FILE $WORK_PATH/




#设置工作目录
WORKDIR $WORK_PATH




# 指定于外界交互的端口
EXPOSE $EXPOSE_PORT
# 配置容器,使其可执行化
ENTRYPOINT exec java $JAVA_OPTS -jar $JAR_FILE

k8s部署模版deployment.yaml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: APP_NAME
  labels:
    app: APP_NAME
spec:
  selector:
    matchLabels:
      app: APP_NAME
  replicas: REPLICAS                            # Pod副本数
  strategy:
    type: RollingUpdate                           # 滚动更新策略
    rollingUpdate:
      maxSurge: 1
      maxUnavailable: 0
  template:
    metadata:
      name: APP_NAME
      labels:
        app: APP_NAME
      annotations:
        timestamp: TIMESTAMP
        prometheus.io/port: "50000"               # 不能动态赋值
        prometheus.io/path: /actuator/prometheus
        prometheus.io/scrape: "true"              # 基于pod的服务发现
    spec:
      affinity:                                   # 设置调度策略,采取多主机/多可用区部署
        podAntiAffinity:
          preferredDuringSchedulingIgnoredDuringExecution:
          - weight: 100
            podAffinityTerm:
              labelSelector:
                matchExpressions:
                - key: app
                  operator: In
                  values:
                  - APP_NAME
              topologyKey: "kubernetes.io/hostname" # 多可用区为"topology.kubernetes.io/zone"
      terminationGracePeriodSeconds: 30             # 优雅终止宽限期
      containers:
      - name: APP_NAME
        image: IMAGE_URL
        imagePullPolicy: Always
        ports:
        - containerPort: APP_PORT
        - name: management-port
          containerPort: 50000         # 应用管理端口
        readinessProbe:                # 就绪探针
          httpGet:
            path: /actuator/health/readiness
            port: management-port
          initialDelaySeconds: 30      # 延迟加载时间
          periodSeconds: 10            # 重试时间间隔
          timeoutSeconds: 1            # 超时时间设置
          successThreshold: 1          # 健康阈值
          failureThreshold: 9          # 不健康阈值
        livenessProbe:                 # 存活探针
          httpGet:
            path: /actuator/health/liveness
            port: management-port
          initialDelaySeconds: 30      # 延迟加载时间
          periodSeconds: 10            # 重试时间间隔
          timeoutSeconds: 1            # 超时时间设置
          successThreshold: 1          # 健康阈值
          failureThreshold: 6          # 不健康阈值
        resources:                     # 容器资源管理
          limits:                      # 资源限制(监控使用情况)
            cpu: 0.5
            memory: 1Gi
          requests:                    # 最小可用资源(灵活调度)
            cpu: 0.1
            memory: 200Mi
        env:
          - name: TZ
            value: Asia/Shanghai
---
kind: HorizontalPodAutoscaler            # 弹性伸缩控制器
apiVersion: autoscaling/v2beta2
metadata:
  name: APP_NAME
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: APP_NAME
  minReplicas: REPLICAS                # 缩放范围
  maxReplicas: 6
  metrics:
    - type: Resource
      resource:
        name: cpu                        # 指定资源指标
        target:
          type: Utilization
          averageUtilization: 50

-End-

最近有一些小伙伴,让我帮忙找一些 面试题 资料,于是我翻遍了收藏的 5T 资料后,汇总整理出来,可以说是程序员面试必备!所有资料都整理到网盘了,欢迎下载!

点击👆卡片,关注后回复【面试题】即可获取

在看点这里好文分享给更多人↓↓

以上是关于K8s + SpringBoot实现零宕机发布的主要内容,如果未能解决你的问题,请参考以下文章

Rancher 2.4实现零宕机升级集群,无需担心组件出现短暂故障!

Tungsten Fabric SDN — 零宕机(ZIU)自动化运维

「故障演练」 Redis Cluster集群,当master宕机,主从切换

K8S架构原理及其工作流程

K8S架构原理及其工作流程

HBase宕机恢复-SplitWAL