1、命名空间管理

命名空间用于集群资源隔离，k8s 内置 default、kube‑system、kube‑public、kube‑node‑lease 等 ns。

# 查看命名空间
kubectl get namespaces
# 创建命名空间
kubectl create namespace timinglee
# 删除命名空间（会删除ns内所有资源）
kubectl delete namespaces timinglee

2、命令式 Pod 基础操作

# 查看pod，-o wide 显示节点、PodIP
kubectl get pods -o wide
# 命令式创建pod
kubectl run lee --image nginx:latest
# 查看pod详细事件（排错核心，ImagePullBackOff镜像拉取失败看这里）
kubectl describe pods error
# 删除单个pod
kubectl delete pods error
# 删除当前命名空间所有pod
kubectl delete pods --all

3、kubectl 常用实操命令

#create：命令式创建deployment
kubectl create deployment webcluster --replicas 2 --image myapp:v1

#edit：直接编辑在线资源yaml
kubectl edit deployments.apps webcluster

#patch：json片段修改资源
kubectl patch deployments.apps webcluster -p '{"spec":{"replicas":1}}'

#expose：创建service暴露应用
kubectl expose deployment webcluster --port 80 --target-port 80

#logs：查看pod容器日志
kubectl logs pods/webcluster-77c87d9946-gh9v7

#attach：连接容器标准输入输出
kubectl attach pods/testpod -it

#exec：进入容器执行命令（最常用）
kubectl exec -it pods/testpod -c testpod -- /bin/bash

#cp：宿主机和pod之间拷贝文件
kubectl cp testpod:/usr/share/nginx/html/index.html  /mnt/test
kubectl cp /mnt/index.html testpod:/usr/share/nginx/html/index.html

#rollout：版本更新、重启、历史
kubectl rollout status deployment webcluster
kubectl rollout restart deployment webcluster
kubectl rollout history deployment webcluster

#scale：扩缩副本数
kubectl scale deployment webcluster --replicas 4

#label：增删标签，`key-`代表删除标签
kubectl label pods webcluster-9787d97f6-zl6q2 app-
kubectl label pods webcluster-9787d97f6-zl6q2 app=webcluster

4、YAML 声明式编写 Pod

Pod 多容器

apiVersion: v1
kind: Pod
metadata:
  labels:
    run: testpod
  name: testpod
spec:
  containers:
  - image: myapp:v1
    name: myapp1
  - image: busyboxplus:latest
    name: busybox
    command:
    - /bin/sh
    - -c
    - sleep 10000

hostPort 端口映射

apiVersion: v1
kind: Pod
metadata:
  labels:
    run: testpod
  name: testpod
spec:
  containers:
  - image: myapp:v1
    name: myapp1
    ports:
    - name: http
      containerPort: 80   #容器内部端口
      hostPort: 80        #宿主机节点端口
      protocol: TCP

env 环境变量注入

apiVersion: v1
kind: Pod
metadata:
  labels:
    run: mysql
  name: mysql
spec:
  containers:
  - image: mysql:8.0
    name: mysql8
    env:
    - name: MYSQL_ROOT_PASSWORD
      value: lee

nodeSelector 指定调度节点

apiVersion: v1
kind: Pod
metadata:
  labels:
    run: mysql
  name: mysql
spec:
  nodeSelector:
    kubernetes.io/hostname: k8s-node2
  containers:
  - image: mysql:8.0
    name: mysql8

hostNetwork 共享宿主机网络栈

apiVersion: v1
kind: Pod
metadata:
  labels:
    run: testpod
  name: testpod
spec:
  hostNetwork: true
  containers:
  - image: busybox:latest
    name: busybox
    command:
    - /bin/sh
    - -c
    - sleep 10000

5、Pod QoS 资源服务质量等级

BestEffort：不配置 requests/limits，优先级最低

resources: {} #不写resources字段

Burstable：requests < limits，优先级中等

resources:
  limits:
    cpu: 700m
    memory: 200M
  requests:
    cpu: 500m
    memory: 100M

Guaranteed：requests 与 limits 完全相等，优先级最高

resources:
  limits:
    cpu: 500m
    memory: 100M
  requests:
    cpu: 500m
    memory: 100M

6、restartPolicy 容器重启策略（Pod 级别）

Always：无论容器怎么退出，总是重启（deployment 默认）

OnFailure：容器异常退出（返回码≠0）才重启；正常退出不重启

Never：无论退出状态，永远不重启（Job 常用）

apiVersion: v1
kind: Pod
metadata:
  labels:
    run: testpod
  name: testpod
spec:
  restartPolicy: OnFailure
  containers:
  - image: busybox:latest
    name: busybox
    command: ["/bin/sh","-c","sleep 30"]

7、Pod 生命周期

init 容器

initContainers：初始化容器，必须全部成功执行完毕后，才启动业务容器，常用于等待依赖、初始化配置。

apiVersion: v1
kind: Pod
metadata:
  labels:
    run: webserver
  name: webserver
spec:
  initContainers:
  - name: busybox
    image: busybox:latest
    command:
    - /bin/sh
    - -c
    - "until test -e /testfile;do echo wating for myservice; sleep 2;done"
  containers:
  - image: myapp:v1
    name: webserver
  restartPolicy: Always

livenessProbe 存活探针

检测容器是否正常运行；探测失败，k8s重启容器。支持 tcpSocket、httpGet、exec。

containers:
- image: myapp:v1
  name: testpod
  livenessProbe:
    tcpSocket:
      port: 80
    initialDelaySeconds: 3 #启动后多久第一次探测
    periodSeconds: 1        #探测间隔
    timeoutSeconds: 1       #探测超时

readinessProbe 就绪探针

检测业务是否就绪；探测失败，不会重启容器，会把 pod 从 service 的 endpoint 摘除，不再接收流量。

containers:
- image: myapp:v1
  name: webserver
  readinessProbe:
    httpGet:
      path: /index.html
      port: 80
    initialDelaySeconds: 3
    periodSeconds: 2
    timeoutSeconds: 1

8、Deployment 版本升级与回退

#升级镜像
kubectl set image deployments webcluster myapp=myapp:v2
#记录更新说明，写入revision
kubectl annotate deployment webcluster kubernetes.io/change-cause="myappv2" --overwrite
#查看版本历史
kubectl rollout history deployment webcluster
#回退到指定版本
kubectl rollout undo deployment webcluster --to-revision=1
#暂停更新
kubectl rollout pause deployment webcluster
#恢复更新
kubectl rollout resume deployment webcluster