Rolling release, graceful downtime, elastic scaling, application monitoring, configuration separation in Spring Boot + K8S

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

K8s + SpringBoot achieves zero downtime release: health check + rolling update + graceful shutdown + elastic scaling + Prometheus monitoring + configuration separation (mirror reuse)

configuration

health examination

Health check type: readiness probe (readiness) + survival probe (liveness)
Probe type: exec (enter the container to execute the script), tcpSocket (detect port), httpGet (call interface)

business level

The project depends on pom.xml

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

Define access ports, paths and permissions 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

Two interfaces will be exposed , /actuator/health/readinessand the access methods are as follows:/actuator/health/liveness

http://127.0.0.1:50000/actuator/health/readiness
http://127.0.0.1:50000/actuator/health/liveness

Operation and maintenance level

k8s deployment template 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

rolling update

The rolling update strategy of k8s resource scheduling, in order to achieve zero downtime release, needs to support health check

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

graceful shutdown

In K8s, before we implement rolling upgrades, we must implement graceful shutdown at the application level. Otherwise, the rolling upgrade will still affect the business. Make the application close the thread, release the connection resource and then stop the service

business level

The project depends on pom.xml

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

Define access ports, paths and permissions 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

The interface will be exposed /actuator/shutdown, and the calling method is as follows:

curl -X POST 127.0.0.1:50000/actuator/shutdown

Operation and maintenance level

Make sure that the dockerfile template integrates the curl tool, otherwise the curl command cannot be used

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 deployment template deployment.yaml

Note: It has been verified that the java project can omit the configuration of the end callback hook

In addition, if you need to use the callback hook, you need to ensure that the curl tool is included in the image, and you need to pay attention to the fact that the application management port (50000) cannot be exposed to the public network

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"]

elastic expansion

After setting resource limits for the pods, create the 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 integration

business level

The project depends on 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>

Define access ports, paths and permissions 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

Will expose /actuator/metricand /actuator/prometheusinterface, the access method is as follows:

http://127.0.0.1:50000/actuator/metric
http://127.0.0.1:50000/actuator/prometheus

Operation and maintenance level

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

configuration separation

Solution: Mount the external configuration file through configmap, and specify the activation environment to run

Role: Separation of configurations to avoid leakage of sensitive information; multiplexing of images to improve delivery efficiency

Generate configmap from file

# 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

Mount configmap and specify the activation environment

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}

summary configuration

business level

The project depends on 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>

Define access ports, paths and permissions 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

Operation and maintenance level

Make sure that the dockerfile template integrates the curl tool, otherwise the curl command cannot be used

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 deployment template 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

------

We have created a high-quality technical exchange group. When you are with excellent people, you will become excellent yourself. Hurry up and click to join the group and enjoy the joy of growing together. In addition, if you want to change jobs recently, I spent 2 weeks a year ago collecting a wave of face-to-face experience from big factories. If you plan to change jobs after the festival, you can click here to claim it !