Azure Kubernetes集群中Spring Boot Admin Server无法注册新Pod

问题描述 #

在Azure Kubernetes集群(AKS)上部署Spring Boot Admin后，初始状态下应用Pod显示为UP；但重启Pod的Deployment后，所有Pod显示为Down，新Pod无法被Spring Boot Admin Server识别，旧Pod仍滞留在控制台中。

用户提供的配置与代码信息：

应用配置（application.yml） #

spring:
  boot:
    admin:
      discovery:
        enabled: true
      monitor:
        status-interval: 3s
        status-lifetime: 3s
  application:
    name: monitoring-server
  cloud:
    kubernetes:
      discovery:
        primary-port-name: health
        enabled: true
        service-labels:
          spring.boot: middleware
      enabled: true

主类代码 #

@SpringBootApplication
@EnableAdminServer
@EnableDiscoveryClient
@EnableScheduling
public class MonitoringApplication {

    public static void main(String[] args) {
        SpringApplication.run(MonitoringApplication.class, args);
    }
}

依赖配置 #

<dependency>
            <groupId>de.codecentric</groupId>
            <artifactId>spring-boot-admin-starter-server</artifactId>
        </dependency>
        <dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-starter-actuator</artifactId>
        </dependency>
        <dependency>
            <groupId>org.springframework.cloud</groupId>
            <artifactId>spring-cloud-starter-kubernetes-client-all</artifactId>
        </dependency>
        <dependency>
            <groupId>de.codecentric</groupId>
            <artifactId>spring-boot-admin-server-ui</artifactId>
        </dependency>

版本信息 #

spring boot version: 2.6.6
spring boot admin version: 2.6.3
spring cloud version: 2021.0.1

排查与解决方案 #

1. 优化Kubernetes服务发现的端点刷新机制 #

默认情况下，Spring Cloud Kubernetes的端点刷新间隔较长，导致SBA无法及时感知Pod的替换。修改配置增加端点刷新频率：

spring:
  cloud:
    kubernetes:
      discovery:
        primary-port-name: health
        enabled: true
        service-labels:
          spring.boot: middleware
        # 开启服务目录监控并设置刷新间隔
        catalog-services-watch:
          enabled: true
          delay: 5000 # 每5秒刷新一次服务端点
      enabled: true

2. 调整Spring Boot Admin的监控与实例管理配置 #

修改SBA的配置，优化实例的状态判断和自动注销逻辑：

spring:
  boot:
    admin:
      discovery:
        enabled: true
        auto-deregister: true # 自动注销下线的实例
      monitor:
        status-interval: 3s
        status-lifetime: 10s # 延长状态有效期，避免短时间断连误判Down
        default-retries: 2 # 增加重试次数，提升状态判断准确性

3. 确保被监控应用的Actuator端点正确暴露 #

被监控的应用必须正确暴露Actuator端点，否则SBA无法获取健康状态：

# 被监控应用的application.yml配置示例
management:
  endpoints:
    web:
      exposure:
        include: health,info,metrics # 至少暴露health端点
  endpoint:
    health:
      show-details: always

4. 验证Kubernetes Service的端口与标签匹配 #

确保被监控应用的Service配置中，health端口指向Actuator的服务端口，且标签与SBA配置的service-labels一致：

apiVersion: v1
kind: Service
metadata:
  name: your-app-service
  labels:
    spring.boot: middleware # 与SBA配置的service-labels对应
spec:
  ports:
    - name: health
      port: 8080
      targetPort: 8080 # 对应被监控应用的Actuator端口
  selector:
    app: your-app

5. 配置Kubernetes API访问权限 #

确保SBA的Pod拥有足够权限访问Kubernetes API以获取服务和Pod信息，创建对应的Role和RoleBinding：

# Role配置
apiVersion: rbac.authorization.k8s.io/v1
kind: Role
metadata:
  name: sba-discovery-role
rules:
- apiGroups: [""]
  resources: ["services", "endpoints", "pods"]
  verbs: ["get", "list", "watch"]
---
# RoleBinding配置
apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
  name: sba-discovery-binding
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: Role
  name: sba-discovery-role
subjects:
- kind: ServiceAccount
  name: default # 替换为SBA使用的ServiceAccount名称
  namespace: your-namespace # 替换为SBA部署的命名空间

6. 版本兼容性确认 #

当前使用的版本组合（Spring Boot 2.6.6 + Spring Boot Admin 2.6.3 + Spring Cloud 2021.0.1）是兼容的，无需调整版本。

内容的提问来源于stack exchange，提问作者ed40721