20K Java+AI 岗位高频考点 — 容器化和K8s是现代后端的必备技能,面试常考Docker原理、K8s核心概念、CI/CD流水线。覆盖Docker核心、K8s架构、Helm、监控、Java容器化。
三大核心技术:
1. Namespace:资源隔离(PID/NET/MNT/UTS/IPC/USER)
2. Cgroups:资源限制(CPU/内存/IO)
3. UnionFS:镜像分层(overlay2)
镜像分层:
- Base层(如eclipse-temurin:21)
- 依赖层(npm install / maven dependency)
- 应用层(jar文件)
- 每层只存储差异,共享相同层
容器 vs 虚拟机:
| 维度 | 容器 | 虚拟机 |
|------|------|--------|
| 隔离 | 进程级 | OS级 |
| 启动 | 秒级 | 分钟级 |
| 大小 | MB | GB |
| 性能 | 接近原生 | 有损耗 |
# 多阶段构建(减小镜像体积)
FROM eclipse-temurin:21-jdk AS build
WORKDIR /build
COPY pom.xml .
COPY src ./src
RUN --mount=type=cache,target=/root/.m2 \
mvn clean package -DskipTests -B
FROM eclipse-temurin:21-jre-alpine
RUN addgroup -S app && adduser -S app -G app
USER app
COPY --from=build /build/target/*.jar app.jar
EXPOSE 8080
HEALTHCHECK --interval=30s CMD curl -f http://localhost:8080/actuator/health || exit 1
ENTRYPOINT ["java", "-XX:+UseZGC", "-XX:MaxRAMPercentage=75.0", "-jar", "app.jar"]
Java 10+自动感知容器资源限制:
- -XX:MaxRAMPercentage=75.0(推荐,自适应)
- 不再需要-Xmx硬编码
GC选择:
- ZGC:低延迟(推荐)
- G1:通用
- Serial:小容器/Serverless
容器内存 = JVM堆 + 元空间 + 线程栈 + 堆外
建议:JVM堆 ≤ 容器内存的75%
Master节点:
- kube-apiserver:集群入口(REST API)
- etcd:分布式KV存储(集群状态)
- kube-scheduler:Pod调度
- kube-controller-manager:控制器(ReplicaSet/Deployment/Node)
Worker节点:
- kubelet:管理Pod生命周期
- kube-proxy:网络代理(Service)
- Container Runtime:容器运行时(containerd)
工作流程:
kubectl → API Server → etcd(存储期望状态)
→ Scheduler(调度Pod到Node)
→ Kubelet(拉起容器)
→ Controller(维护期望状态)
状态:Pending → Running → Succeeded/Failed
生命周期钩子:
- postStart:容器创建后执行
- preStop:容器终止前执行
健康检查:
- livenessProbe:存活检查(失败→重启容器)
- readinessProbe:就绪检查(失败→从Service摘除)
- startupProbe:启动检查(启动期间覆盖其他探针)
yaml:
livenessProbe:
httpGet: { path: /actuator/health/liveness, port: 8080 }
initialDelaySeconds: 30
periodSeconds: 10
readinessProbe:
httpGet: { path: /actuator/health/readiness, port: 8080 }
initialDelaySeconds: 10
periodSeconds: 5
apiVersion: apps/v1
kind: Deployment
metadata: { name: order-service }
spec:
replicas: 3
strategy:
type: RollingUpdate
rollingUpdate:
maxSurge: 1 # 最多多1个Pod
maxUnavailable: 0 # 不允许不可用
template:
spec:
containers:
- name: order-service
image: registry.com/order:v2
readinessProbe:
httpGet: { path: /actuator/health/readiness, port: 8080 }
terminationGracePeriodSeconds: 60
Service类型:
- ClusterIP:集群内部访问(默认)
- NodePort:节点端口暴露
- LoadBalancer:云厂商负载均衡
- ExternalName:CNAME映射
Ingress(HTTP路由):
apiVersion: networking.k8s.io/v1
kind: Ingress
spec:
rules:
- host: api.example.com
http:
paths:
- path: /orders
backend:
service: { name: order-service, port: { number: 8080 } }
- path: /users
backend:
service: { name: user-service, port: { number: 8080 } }
# ConfigMap(非敏感配置)
apiVersion: v1
kind: ConfigMap
metadata: { name: order-config }
data:
SPRING_PROFILES_ACTIVE: "prod"
JAVA_OPTS: "-XX:+UseZGC -XX:MaxRAMPercentage=75.0"
# Secret(敏感信息,base64编码)
apiVersion: v1
kind: Secret
metadata: { name: db-secret }
type: Opaque
data:
username: YWRtaW4=
password: c2VjcmV0MTIz
# Pod引用
spec:
containers:
- envFrom:
- configMapRef: { name: order-config }
- secretRef: { name: db-secret }
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
spec:
scaleTargetRef:
apiVersion: apps/v1
kind: Deployment
name: order-service
minReplicas: 2
maxReplicas: 10
metrics:
- type: Resource
resource:
name: cpu
target: { type: Utilization, averageUtilization: 70 }
- type: Pods
pods:
metric: { name: http_requests_per_second }
target: { type: AverageValue, averageValue: 1000 }
name: CI/CD
on:
push: { branches: [main] }
pull_request: { branches: [main] }
jobs:
test:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/setup-java@v4
with: { java-version: '21', distribution: 'temurin', cache: 'maven' }
- run: mvn verify -B -T 4
build-push:
needs: test
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- run: |
docker build -t ${{ secrets.REGISTRY }}/order:${{ github.sha }} .
docker push ${{ secrets.REGISTRY }}/order:${{ github.sha }}
deploy:
needs: build-push
runs-on: ubuntu-latest
steps:
- run: |
kubectl set image deployment/order-service \
order=${{ secrets.REGISTRY }}/order:${{ github.sha }}
GitOps原则:
1. Git是唯一真相源
2. 声明式描述系统状态
3. 自动拉取并应用变更
4. 持续协调期望状态和实际状态
ArgoCD流程:
1. Git提交K8s YAML变更
2. ArgoCD检测到变更
3. 自动同步到K8s集群
4. 可配置自动/手动同步
优势:
- Git审计追踪
- 一键回滚(git revert)
- 多集群管理
- 与PR流程集成
蓝绿发布:
- 蓝色(当前版本) + 绿色(新版本) 并行
- 验证通过后切流量到绿色
- 回滚:切回蓝色
金丝雀发布:
- 新版本逐步放量:5% → 20% → 50% → 100%
- 每步监控指标(错误率/延迟)
- 异常自动回滚
K8s实现:
# Istio VirtualService权重路由
apiVersion: networking.istio.io/v1beta1
kind: VirtualService
spec:
http:
- route:
- destination: { host: order-service, subset: v1 }
weight: 90
- destination: { host: order-service, subset: v2 }
weight: 10
Prometheus:
- 时序数据库
- Pull模式拉取指标
- PromQL查询语言
- 告警规则(AlertManager)
Spring Boot Actuator + Micrometer:
management.endpoints.web.exposure.include=health,prometheus
management.metrics.export.prometheus.enabled=true
关键指标:
- http_server_requests_seconds:请求延迟
- jvm_memory_used_bytes:JVM内存
- system_cpu_usage:CPU使用率
- hikaricp_connections_active:连接池
Grafana Dashboard:
- Spring Boot Statistics
- JVM Micrometer
- Kubernetes Pod Monitoring
架构:Fluentd/Fluent Bit → Elasticsearch → Kibana
日志规范:
1. JSON格式(方便解析)
2. 包含traceId(链路追踪)
3. 日志级别合理(ERROR/WARN/INFO/DEBUG)
Spring Boot日志配置:
logging.pattern.console=%d{ISO8601} [%X{traceId}] %-5level %logger - %msg%n
# Spring Boot配置
management:
tracing:
sampling.probability: 1.0
otlp:
tracing:
endpoint: http://otel-collector:4317
# 指标+日志+链路追踪 三合一
# 可对接:Jaeger / Zipkin / Tempo
| 维度 | Docker | Docker Swarm | Kubernetes |
|---|---|---|---|
| 定位 | 容器运行时 | 编排(简单) | 编排(强大) |
| 扩缩容 | 手动 | 简单自动 | HPA/VPA |
| 服务发现 | -- | 内置 | Service+CoreDNS |
| 滚动更新 | -- | 支持 | 完善 |
| 生态 | 大 | 小 | 最大 |
| 适用 | 单机 | 小集群 | 生产环境 |
1. JVM参数:
-XX:+UseContainerSupport(默认开启)
-XX:MaxRAMPercentage=75.0
-XX:+UseZGC
2. 健康检查:
Spring Actuator + K8s Probe
/actuator/health/liveness
/actuator/health/readiness
3. 优雅停机:
server.shutdown=graceful
spring.lifecycle.timeout-per-shutdown-phase=30s
4. 安全:
非root用户运行
只读文件系统
最小基础镜像(alpine)
5. GC调优:
小容器(<1GB):Serial GC
中容器(1-4GB):G1 GC
大容器(>4GB):ZGC
# Pod状态异常
kubectl describe pod <pod-name> # 查看事件
kubectl logs <pod-name> -f # 查看日志
kubectl logs <pod-name> --previous # 上一次容器的日志(崩溃后)
# 常见问题:
# ImagePullBackOff → 镜像拉取失败(检查镜像名/Secret)
# CrashLoopBackOff → 容器启动后崩溃(检查日志)
# Pending → 无法调度(资源不足/标签不匹配)
# OOMKilled → 内存不足(增加resources.limits.memory)
# 节点问题
kubectl describe node <node-name>
kubectl top pods # 查看资源使用
1. 资源限制:每个容器必须设置requests和limits
2. 健康检查:livenessProbe + readinessProbe
3. 亲和性:Pod反亲和(分散部署)
4. PDB:PodDisruptionBudget(保证最少可用副本)
5. 网络策略:NetworkPolicy限制Pod间通信
6. RBAC:最小权限原则
7. Secret管理:External Secrets / Vault
8. 镜像安全:Trivy扫描漏洞
Docker三核心:Namespace隔离、Cgroups限制、UnionFS分层
>
K8s四组件:API Server(入口)、etcd(存储)、Scheduler(调度)、Kubelet(执行)
>
发布三策略:蓝绿(全量切换)、金丝雀(逐步放量)、滚动(逐个更新)
>
可观测三柱:日志(EFK)、指标(Prometheus)、链路(OpenTelemetry)