K8s学习笔记：02-Deployment控制器

目标：理解为什么需要 Deployment，掌握创建、扩缩容、滚动更新和回滚

---

2.1 Pod 的局限与 Deployment 的诞生

回顾：Pod 能干什么

第一章你学会了用 kubectl run 创建 Pod。Pod 能跑起来，但它有个致命问题：

kubectl delete pod my-nginx
# Pod 删了就是删了，没人会帮你重建

Pod 是”宠物”，不是”牛”——死了就是死了。

真实需求

需求	Pod 能做到吗？
挂了自动重启	✗
想跑 3 个副本	✗
更新镜像不停机	✗
版本出问题回滚	✗

Deployment 出场

┌────────── Deployment ───────────┐
│  replicas: 3                    │
│  image: nginx:alpine            │
│  ┌──────── ReplicaSet ────────┐ │
│  │  Pod-abc   Pod-def  Pod-ghi│ │
│  └────────────────────────────┘ │
└─────────────────────────────────┘

Deployment 是 Pod 的”管理者”。你告诉它”我想要 3 个 nginx”，它通过 ReplicaSet 确保始终有 3 个 Pod 在跑。Pod 挂了？自动补一个。想更新镜像？一个一个替换，保证服务不中断。

类比 Docker：

Docker	K8s
docker run	Pod
docker-compose up	Deployment

---

2.2 创建与管理 Deployment

创建第一个 Deployment

kubectl create deployment my-nginx-deploy --image=nginx:alpine --port=80

查看 Deployment

kubectl get deploy
# NAME              READY   UP-TO-DATE   AVAILABLE   AGE
# my-nginx-deploy   1/1     1            1           10s

kubectl get rs
# NAME                         DESIRED   CURRENT   READY   AGE
# my-nginx-deploy-xxxxxxxxxx   1         1         1       10s

kubectl get pods
# NAME                               READY   STATUS    RESTARTS   AGE
# my-nginx-deploy-xxxxxxxxxx-xxxxx   1/1     Running   0          10s

注意 Pod 名字的变化：不再是 my-nginx，而是 my-nginx-deploy-<rs-hash>-<pod-hash>。Deployment 创建了 ReplicaSet，ReplicaSet 创建了 Pod。三层关系：

Deployment → ReplicaSet → Pod

查看详细信息

kubectl describe deploy my-nginx-deploy

root@bote798-ubuntu:~# kubectl describe deploy my-nginx-deploy
Name:                   my-nginx-deploy
Namespace:              default
CreationTimestamp:      Mon, 13 Jul 2026 20:32:24 +0800
Labels:                 app=my-nginx-deploy
Annotations:            deployment.kubernetes.io/revision: 1
Selector:               app=my-nginx-deploy
Replicas:               1 desired | 1 updated | 1 total | 1 available | 0 unavailable
StrategyType:           RollingUpdate
MinReadySeconds:        0
RollingUpdateStrategy:  25% max unavailable, 25% max surge
Pod Template:
  Labels:  app=my-nginx-deploy
  Containers:
   nginx:
    Image:         nginx:alpine
    Port:          80/TCP
    Host Port:     0/TCP
    Environment:   <none>
    Mounts:        <none>
  Volumes:         <none>
  Node-Selectors:  <none>
  Tolerations:     <none>
Conditions:
  Type           Status  Reason
  ----           ------  ------
  Available      True    MinimumReplicasAvailable
  Progressing    True    NewReplicaSetAvailable
OldReplicaSets:  <none>
NewReplicaSet:   my-nginx-deploy-5d4cc55964 (1/1 replicas created)
Events:
  Type    Reason             Age    From                   Message
  ----    ------             ----   ----                   -------
  Normal  ScalingReplicaSet  6m10s  deployment-controller  Scaled up replica set my-nginx-deploy-5d4cc55964 to 1

关注输出里的 Conditions 和 Events，和 Pod 的 describe 类似但更宏观。

测试自愈能力

# 记住当前 Pod 的名字
kubectl get pods

NAME                               READY   STATUS    RESTARTS   AGE
my-nginx-deploy-5d4cc55964-zjl9d   1/1     Running   0          7m29s

# 删掉它
kubectl delete pod <pod-name>

root@bote798-ubuntu:~# kubectl delete pod my-nginx-deploy-5d4cc55964-zjl9d
pod "my-nginx-deploy-5d4cc55964-zjl9d" deleted from default namespace

# 立刻再查——一个新 Pod 已经起来了
kubectl get pods

NAME                               READY   STATUS    RESTARTS   AGE
my-nginx-deploy-5d4cc55964-n5p7g   1/1     Running   0          5s

Deployment 发现 Pod 少了，立刻让 ReplicaSet 补了一个。这就是”声明式”：你声明了目标状态（1 个 Pod），K8s 持续确保实际状态 = 目标状态。

扩缩容

# 扩容到 3 个副本
kubectl scale deploy my-nginx-deploy --replicas=3

kubectl get pods -o wide

NAME                               READY   STATUS    RESTARTS   AGE    IP           NODE                 NOMINATED NODE   READINESS GATES
my-nginx-deploy-5d4cc55964-n5p7g   1/1     Running   0          113s   10.244.0.6   kind-control-plane   <none>           <none>
my-nginx-deploy-5d4cc55964-xrd9n   1/1     Running   0          11s    10.244.0.7   kind-control-plane   <none>           <none>
my-nginx-deploy-5d4cc55964-zv5cg   1/1     Running   0          11s    10.244.0.8   kind-control-plane   <none>           <none>
# 三个 Pod，分布在节点上

# 缩容回 1 个
kubectl scale deploy my-nginx-deploy --replicas=1

kubectl get pods

NAME                               READY   STATUS    RESTARTS   AGE
my-nginx-deploy-5d4cc55964-n5p7g   1/1     Running   0          5m21s
# 只剩一个了

---

2.3 滚动更新与回滚

更新镜像

# 查看当前镜像
kubectl describe deploy my-nginx-deploy | grep Image

# 更新镜像版本（nginx:alpine → nginx:alpine-perl）
kubectl set image deploy/my-nginx-deploy nginx=nginx:alpine-perl

# 实时观察更新过程
kubectl rollout status deploy/my-nginx-deploy

Waiting for deployment "my-nginx-deploy" rollout to finish: 1 old replicas are pending termination...
Waiting for deployment "my-nginx-deploy" rollout to finish: 1 old replicas are pending termination...
deployment "my-nginx-deploy" successfully rolled out

滚动更新的过程：

旧 Pod (running) ──→ 新 Pod (creating) ──→ 新 Pod (running) ──→ 旧 Pod (terminating)

每次只替换一个，保证服务不中断。

查看更新历史

kubectl rollout history deploy/my-nginx-deploy
# REVISION  CHANGE-CAUSE
# 1         <none>
# 2         <none>

回滚

# 回滚到上一个版本
kubectl rollout undo deploy/my-nginx-deploy

# 或者回滚到指定版本
kubectl rollout undo deploy/my-nginx-deploy --to-revision=1

# 确认回滚成功
kubectl describe deploy my-nginx-deploy | grep Image

---

2.4 章节练习

练习 1：自愈测试

# 1. 创建 Deployment，3 个副本，使用 nginx:alpine
# 2. 用 kubectl get pods -o wide 查看 3 个 Pod
# 3. 手动删除其中一个 Pod
# 4. 立刻 kubectl get pods——观察 Pod 数量变化
# 5. 再次删除一个 Pod，记下新旧 Pod 的名字变化

练习 2：扩缩容

# 1. 把副本数从 3 扩到 5
# 2. kubectl get pods 确认 5 个都在运行
# 3. 缩到 2
# 4. 确认多余的 3 个被终止

练习 3：滚动更新与回滚

# 1. 把镜像更新为 nginx:alpine-perl
# 2. 用 rollout status 观察更新过程
# 3. 确认更新完成（kubectl describe deploy | grep Image）
# 4. 查看 rollout history
# 5. 回滚到旧版本
# 6. 确认镜像已回滚

练习 4：模拟故障回滚

# 1. 故意更新为一个不存在的镜像：nginx:not-exist
# 2. kubectl get pods 观察状态（会看到 ImagePullBackOff 或 CrashLoopBackOff）
# 3. kubectl rollout status 会一直等待——Ctrl+C 退出
# 4. kubectl rollout undo 立即回滚
# 5. kubectl get pods——恢复正常

自检清单

• 理解 Pod vs Deployment 的区别和关系
• 能看懂 Deployment → ReplicaSet → Pod 的层级
• 能手操 scale 扩缩容
• 能手操 set image + rollout undo 更新和回滚
• 见识过自愈：删 Pod 自动补回
• 见识过坏更新 + 快速回滚恢复