antony@notes:~/kubernetes$ cat "Kubernetes-Deployment.md"

Kubernetes Deployment

2023-02-08· kubernetes ·系統工程

Kubernetes Deployment

前言

本篇主要介紹什麼是 Deployment object，並且透過這個 K8S 的物件，將一個網站服務進行進版或退版，且服務不中斷。

認識 Deployment

A Deployment provides declarative updates for Pods and ReplicaSets.

You describe a desired state in a Deployment, and the Deployment Controller changes the actual state to the desired state at a controlled rate. You can define Deployments to create new ReplicaSets, or to remove existing Deployments and adopt all their resources with new Deployments.

一個在 etcd 中的控制區塊，裡面紀錄 ReplicaSet Controller 這個功能
Deployment 由 ReplicaSet 和 Pods 組成
Deployment 管理 ReplicaSet ， ReplicaSet 管理 Pods
Deployment 可做到 Replication of components + Auto-scaling + Load balancing + Rolling updates
Rolling updates，是由 Deployment 在做的
讓企業 run 的應用系統能夠長長久久穩穩當當

認識 ReplicaSet

ReplicaSet 是用來確保在 k8s 中，在資源允許的前提下，指定的 pod 的數量會跟使用者所期望的一致，也就是所謂的 desired status。
ReplicaSets 可以生成多個 pod (Scalability，橫向擴增或縮減)，主要目的是服務流量分散
- 多個 pod 是一樣的 (本尊和分身的概念)
- 所有 Pod 都共享同一 PersistentVolumeClaim，並與同一 PersistentVolume 綁定
ReplicaSet 有自我療傷的功能
- 當管理的 pod 毀損或被刪除時，會自動再產生同樣的 pod 出來
- 只要啟動 PV ，pod 的資料就不會不見
可控管 pod 的狀態，如果 pod 負載太重，會變成離線的狀態，處理完手上的事情再回來

撰寫 Depolyment Object 部署檔案

echo 'apiVersion: apps/v1
kind: Deployment
metadata:
  name: depobj
  labels:
    app: deppod
spec:
  replicas: 2
  selector:
    matchLabels:
      app: deppod
  template:
    metadata:
      labels:
        app: deppod
    spec:
      containers:
      - name: myalpine
        image: quay.io/ict39/alpine
        imagePullPolicy: IfNotPresent
        tty: true '> ~/wulin/yaml/depobj.yaml

kind: Deployment，生成 Deployment 的物件
- metadata.name，Deployment 的名字
- labels:，貼標籤，key 是 app，value 是 deppod
spec，Deployment 的規格
- replicas: 2，讓 ReplicaSet 產生 2 個 pod
- selector.matchLabels:，透過標籤管理 pod
  - app: deppod，標籤
- template:，對 pod 的宣告
  - metadata:，pod 的重要資訊
    - labels.app，對 pod 貼 app: deppod 的標籤
  - spec，pod 的規格
    - containers，Container 的設定
      - name，Container 的名字
      - image，Container 用的 image
      - imagePullPolicy: IfNotPresent，如果 node 主機沒有 image 就上網下載
      - tty: true，因 image 內定執行的命令是貝殼程式，所以要給終端機

建立與檢視 Depolyment Object

$ kubectl apply -f wulin/yaml/depobj.yaml
deployment.apps/depobj created

$ kubectl get all --selector app=deppod
NAME                         READY   STATUS    RESTARTS   AGE
pod/depobj-7cb97745c-4k8jm   1/1     Running   0          2m32s
pod/depobj-7cb97745c-ssbf4   1/1     Running   0          2m32s

NAME                 TYPE        CLUSTER-IP   EXTERNAL-IP   PORT(S)   AGE
service/kubernetes   ClusterIP   10.98.0.1    <none>        443/TCP   24h

NAME                     READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/depobj   2/2     2            2           2m32s

NAME                               DESIRED   CURRENT   READY   AGE
replicaset.apps/depobj-7cb97745c   2         2         2       2m32s

先看 Deployment 的名字， depobj
再看 Replicaset 的名字，depobj-7cb97745c
- 因為是由 Deployment 建立的，所以 Replicaset 會有 deplopyment 的名字
再看 pod 的名字，depobj-7cb97745c-4k8jm
- 因為 pod 是由 Deploy 的 ReplicaSet 建立的，所以 pod 名字前面會帶有他們的名字

測試 ReplicaSets Controller

刪除帶有 app=deppod 標籤的 pod

$ kubectl delete pod --selector app=deppod
pod "depobj-7cb97745c-4k8jm" deleted
pod "depobj-7cb97745c-ssbf4" deleted

ReplicaSets Controller 會一直執行 reconciliation loop 程序, 確保 Deployment Object 的 actual state 與 desired state 一致. 上面命令刪除二個 POD, 這時 reconciliation loop 程序會再生成二個新的 POD, 由以下命令得知

$ kubectl get all --selector app=deppod
NAME                         READY   STATUS    RESTARTS   AGE
pod/depobj-7cb97745c-jg7fk   1/1     Running   0          2m23s
pod/depobj-7cb97745c-smjlt   1/1     Running   0          2m23s

NAME                     READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/depobj   2/2     2            2           22m

NAME                               DESIRED   CURRENT   READY   AGE
replicaset.apps/depobj-7cb97745c   2         2         2       22m

pod 的名字後面流水號不同，代表 pod 真的有被重新建立

修改 Depolyment Object 部署檔案

$ echo 'apiVersion: apps/v1
kind: Deployment
metadata:
  name: depobj
  labels:
    app: deppod
spec:
  replicas: 1
  selector:
    matchLabels:
      app: deppod
  template:
    metadata:
      labels:
        app: deppod
    spec:
      containers:
      - name: myderby
        image: quay.io/ict39/alpine.derby
        imagePullPolicy: IfNotPresent
        ports:
        - name: http
          containerPort: 8888 '> ~/wulin/yaml/depobj.yaml

不刪除原本的 Deployment Object，直接再次產生

$ kubectl apply -f wulin/yaml/depobj.yaml
deployment.apps/depobj configured

檢視所有帶有 app=deppod 這個標籤的物件

$ kubectl get all --selector app=deppod
NAME                          READY   STATUS    RESTARTS   AGE
pod/depobj-85987557b8-pq6f7   1/1     Running   0          47s

NAME                     READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/depobj   1/1     1            1           32m

NAME                                DESIRED   CURRENT   READY   AGE
replicaset.apps/depobj-7cb97745c    0         0         0       32m
replicaset.apps/depobj-85987557b8   1         1         1       47s

replicaset.apps/depobj-7cb97745c，注意，這是上一個 Deployment 產生的 ReplicaSet
當 Deployment 裡面的 pod 換成產生不同的 Container ，原本的 pod 會被刪除，但是 ReplicaSet 不會被刪除

建立與檢視 Depolyment Object

echo '
apiVersion: apps/v1
kind: Deployment
metadata:
  name: depng
  labels:
    app: nginx
spec:
  replicas: 2
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
      - name: nginx
        image: quay.io/cloudwalker/nginx:1.20
        ports:
        - containerPort: 80 '> ~/wulin/yaml/depng.yaml

$ kubectl apply -f ~/wulin/yaml/depng.yaml 
deployment.apps/depng created

$ kg all --selector app=nginx
NAME                        READY   STATUS    RESTARTS   AGE
pod/depng-7bff78c9c-ntc7t   1/1     Running   0          2m
pod/depng-7bff78c9c-skj6f   1/1     Running   0          2m

NAME                    READY   UP-TO-DATE   AVAILABLE   AGE
deployment.apps/depng   2/2     2            2           2m

NAME                              DESIRED   CURRENT   READY   AGE
replicaset.apps/depng-7bff78c9c   2         2         2       2m

Updating a Deployment

$ kubectl set image deployment.v1.apps/depng nginx=quay.io/cloudwalker/nginx:1.21 --record 
deployment.apps/depng image updated

將 image 換裝成新版 1.21
--record ，記錄

$ kubectl rollout status deployment.v1.apps/depng
Waiting for deployment "depng" rollout to finish: 1 out of 2 new replicas have been updated...
Waiting for deployment "depng" rollout to finish: 1 out of 2 new replicas have been updated...
Waiting for deployment "depng" rollout to finish: 1 out of 2 new replicas have been updated...
Waiting for deployment "depng" rollout to finish: 1 old replicas are pending termination...
Waiting for deployment "depng" rollout to finish: 1 old replicas are pending termination...
deployment "depng" successfully rolled out

這命令要再上一個命令做完後趕快執行，不然只會看到最後一行
Deplyment 進版時，會先把新的 pod 建立起來，再把舊版本的 pod 刪除

檢查 image 是不是真的換裝城新版的

$ kubectl describe deployments depng
Name:                   dep1
............
Pod Template:
  Labels:  app=nginx
  Containers:
   nginx:
    Image:        quay.io/cloudwalker/nginx:1.21
    Port:         80/TCP
    Host Port:    0/TCP

Rolling Back a Deployment

將 Deployment 退回上一個版本

$ kubectl rollout undo deployment.v1.apps/depng --to-revision=1

檢查是否符合預期

$ kubectl describe deployment depng
........
  Containers:
   nginx:
    Image:        quay.io/cloudwalker/nginx:1.20
    Port:         80/TCP
    Host Port:    0/TCP

Scaling a Deployment

將 Deployment 的 pod 擴充為 3 個

$ kubectl scale deployment.v1.apps/depng --replicas=3

檢查是否符合預期

$ kubectl get deployment depng
NAME   DESIRED   CURRENT   UP-TO-DATE   AVAILABLE   AGE
dep1   3         3         3            3           46m

$ kubectl get pods | grep depng
dep1-66f7f56f56-2rf8p   1/1     Running   0          39s
dep1-66f7f56f56-bggcn   1/1     Running   0          71s
dep1-66f7f56f56-mn589   1/1     Running   0          69s

$ kubectl delete deployment depng

明城老師版本

建立 Local Persistent Volume

$ echo 'apiVersion: v1
kind: PersistentVolume
metadata:
  name: pv-rwo-5m
spec:
  capacity:
    storage: 5Mi
  accessModes:
    - ReadWriteOnce
  local:
    path: "/opt/pv/5m"
  nodeAffinity:
    required:
      nodeSelectorTerms:
      - matchExpressions:
        - key: kubernetes.io/hostname
          operator: In
          values:
          - w1' > pv-rwo-5m.yaml

$ kubectl apply -f pv-rwo-5m.yaml

$ kubectl get pv pv-rwo-5m

建立 Local PersistentVolumeClaim

$ echo 'apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: pvc-rwo-3m
spec:
  accessModes:
    - ReadWriteOnce
  resources:
    requests:
      storage: 3Mi' > pvc-rwo-3m.yaml

$ kubectl apply -f pvc-rwo-3m.yaml

檢視 PV 與 PVC

$ kubectl get pvc pvc-rwo-3m
NAME         STATUS   VOLUME      CAPACITY   ACCESS MODES   STORAGECLASS   AGE
pvc-rwo-3m   Bound    pv-rwo-5m   5Mi        RWO                           10s

bigred@m1:~/0613$ kubectl get pv pv-rwo-5m
NAME        CAPACITY   ACCESS MODES   RECLAIM POLICY   STATUS   CLAIM                STORAGECLASS   REASON   AGE
pv-rwo-5m   5Mi        RWO            Retain           Bound    default/pvc-rwo-3m                           55s

建立 Deployment

$ echo 'apiVersion: apps/v1
kind: Deployment
metadata:
  name: nginx
spec:
  replicas: 3
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
        - name: nginx
          image: quay.io/flysangel/nginx
          volumeMounts:
            - name: html-storage
              mountPath: /usr/share/nginx/html
      volumes:
        - name: html-storage
          persistentVolumeClaim:
            claimName: pvc-rwo-3m' > deploy-nginx.yaml

建立 Pod 使用 PVC

$ ssh w1 sudo mkdir -p /opt/pv/5m
$ ssh w1 "echo 'Hello My Pvc' | sudo tee /opt/pv/5m/index.html"

驗證 nginx 首頁

$ kubectl apply -f deploy-nginx.yaml

$ kubectl get pod -o wide
NAME                     READY   STATUS    RESTARTS   AGE     IP             NODE   NOMINATED NODE   READINESS GATES
nginx-5bfb7cc7b7-2jkbl   1/1     Running   0          2m50s   10.244.1.116   w1     <none>           <none>
nginx-5bfb7cc7b7-9zrbt   1/1     Running   0          2m50s   10.244.1.115   w1     <none>           <none>
nginx-5bfb7cc7b7-g2glm   1/1     Running   0          2m50s   10.244.1.117   w1     <none>           <none>

$ curl http://10.244.1.116
Hello My Pvc

$ curl http://10.244.1.115
Hello My Pvc

$ curl http://10.244.1.117
Hello My Pvc

透過 Deployment/ReplicaSets 建立的 pods 後面會有一串代碼

認識 ReplicaSet

對於 ReplicaSet 所管理的 Pod, 雖然每個 Pod 都有 hostname, 但當 Pod 被移轉時, hostname 會重新分配, 對於指向 ReplicaSet 的 Service, 其他服務訪問 Service 時會被隨機分配到某個 Pod
DNS 伺服器，幫我們把記不住的 IP 轉成記得住的網址
Service 建立後，會跟 K8s core-dns 註冊

建立 Service Deployment

$ echo 'apiVersion: v1
kind: Service
metadata:
  name: nginx
spec:
  clusterIP: None
  selector:
    app: nginx' > svc-nginx.yaml

上面這個 Service 是 Headless Service （無頭服務）。

$ kubectl apply -f svc-nginx.yaml
service/nginx created

$ kubectl get service
NAME         TYPE        CLUSTER-IP   EXTERNAL-IP   PORT(S)   AGE
kubernetes   ClusterIP   10.98.0.1    <none>        443/TCP   13d
nginx        ClusterIP   None         <none>        <none>    14s

觀察 Service Deployment

$ kubectl get service -n kube-system
NAME       TYPE        CLUSTER-IP   EXTERNAL-IP   PORT(S)                  AGE
kube-dns   ClusterIP   10.98.0.10   <none>        53/UDP,53/TCP,9153/TCP   13d

透過 nslookup 來將

$ nslookup
> server 10.98.0.10
Default server: 10.98.0.10
Address: 10.98.0.10#53

> 10.244.1.115
115.1.244.10.in-addr.arpa       name = 10-244-1-115.nginx.default.svc.k8s.org.

> 10.244.1.116
116.1.244.10.in-addr.arpa       name = 10-244-1-116.nginx.default.svc.k8s.org.

> 10.244.1.117
117.1.244.10.in-addr.arpa       name = 10-244-1-117.nginx.default.svc.k8s.org.

> nginx.default.svc.k8s.org
Server:         10.98.0.10
Address:        10.98.0.10#53

Name:   nginx.default.svc.k8s.org
Address: 10.244.1.116
Name:   nginx.default.svc.k8s.org
Address: 10.244.1.117
Name:   nginx.default.svc.k8s.org
Address: 10.244.1.115

練習

請問如何擴展 Deployment。請觀察擴展 Deployment 時，pod 或 container 狀態為何。請將 Local Persistent Volumes 換成 Local Path Provisioner。

練習完後請刪除練習環境 Pod -> PVC

$ kubectl scale deploy/nginx --replicas=6

tags: `系統工程`

Kubernetes Landlord Kubernetes Deployment

Kubernetes Deployment

Kubernetes Deployment

前言

目錄

認識 Deployment

認識 ReplicaSet

撰寫 Depolyment Object 部署檔案

建立與檢視 Depolyment Object

測試 ReplicaSets Controller

修改 Depolyment Object 部署檔案

建立與檢視 Depolyment Object

Updating a Deployment

Rolling Back a Deployment

Scaling a Deployment

明城老師版本

建立 Local Persistent Volume

建立 Local PersistentVolumeClaim

檢視 PV 與 PVC

建立 Deployment

建立 Pod 使用 PVC

驗證 nginx 首頁

認識 ReplicaSet

建立 Service Deployment

觀察 Service Deployment

練習

tags: 系統工程

tags: `系統工程`