學習K8S之路.14---K8S HPA自動進行擴容，縮容

上章，講述了在K8S中部署metrics server，為本章的HPA做了一定的鋪墊。下面簡單說說HPA

HPA全稱是Horizontal Pod Autoscaler，翻譯成中文是POD水平自動伸縮，以下都會用HPA代替Horizontal Pod Autoscaler，HPA可以基於CPU利用率對replication controller、deployment和replicaset中的pod數量進行自動擴縮容（除了CPU利用率也可以基於其他應程序提供的度量指標custom metrics進行自動擴縮容）。pod自動縮放不適用於無法縮放的對象，比如DaemonSets。HPA由Kubernetes API資源和控制器實現。

一：HPA工作原理

      HPA的實現是一個控制循環，由controller manager的–horizontal-pod-autoscaler-sync-period參數指定周期（默認值為15秒）。每個周期內，controller manager根據每個HorizontalPodAutoscaler定義中指定的指標查詢資源利用率。controller manager可以從resource metrics API（pod 資源指標）和custom metrics API（自定義指標）獲取指標。
      官網參考文檔：

https://v1-17.docs.kubernetes.io/zh/docs/tasks/run-application/horizontal-pod-autoscale-walkthrough/

二：HPA API對象

HPA的API有三個版本，通過kubectl api-versions | grep autoscal可看到：
autoscaling/v1
autoscaling/v2beta1
autoscaling/v2beta2

注釋：
autoscaling/v1只支持基於CPU指標的縮放；
autoscaling/v2beta1支持Resource Metrics（資源指標，如pod的CPU）和Custom Metrics（自定義指標）的縮放；
autoscaling/v2beta2支持Resource Metrics（資源指標，如pod的CPU）和Custom Metrics（自定義指標）和ExternalMetrics（額外指標）的縮放

三：基於CPU自動擴容，進行講述：

1：以tomcat為例，創建資源配置清單

-6-96 tomcat-hpa]# vi dp.yaml 
kind: Deployment
metadata:
  name: tomcat-hpa
  namespace: app
spec:
  selector:
    matchLabels:
      app: tomcat-hpa
  template:
    metadata:
      labels:
        app: tomcat-hpa
    spec:
      containers:
      - name: tomcat-hpa
        image: harbor.auth.com/base/tomcat:v1.0
        imagePullPolicy: IfNotPresent
        ports:
        - containerPort: 8080
        resources:
          limits:
            cpu: 200m
          requests:
            cpu: 80m


-6-96 tomcat-hpa]# cat svc.yaml 
apiVersion: v1
kind: Service
metadata:
  name: tomcat-hpa-svc
  namespace: app
spec:
  ports:
  - port: 8080
    protocol: TCP
  selector:
    app: tomcat-hpa
  type: ClusterIP

 2：應用資源配置清單

k8s-6-94 ~]# kubectl apply -f http://k8s-yaml.auth.com/tomcat-hpa/dp.yaml
k8s-6-94 ~]# kubectl apply -f http://k8s-yaml.auth.com/tomcat-hpa/svc.yaml

3：創建HPA

tomcat-hpa服務正在運行，使用kubectl autoscale創建自動縮放器，實現對tomcat-hpa這個deployment創建的pod自動擴縮容，下面的命令將會創建一個HPA，HPA將會根據CPU，內存等資源指標增加或減少副本數，創建一個可以實現如下目的的hpa：

1）讓副本數維持在1-10個之間（這里副本數指的是通過deployment部署的pod的副本數）
2）將所有Pod的平均CPU使用率維持在50％（通過kubectlrun運行的每個pod如果是80毫核，這意味着平均CPU利用率為40毫核

給上面tomcat-hpa這個deployment創建HPA

k8s-6-94 ~]# kubectl autoscale deployment tomcat-hpa -n app --cpu-percent=50 --min=1 --max=10

注釋：

kubectl autoscale deployment tomcat-hpa （tomcat-hpa表示deployment的名字） --cpu-percent=50（表示cpu使用率不超過50%） --min=1（最少一個pod） --max=10（最多10個pod)

4：驗證HPA

-6-95 ~]# kubectl get hpa -n app
NAME         REFERENCE               TARGETS   MINPODS   MAXPODS   REPLICAS   AGE
tomcat-hpa   Deployment/tomcat-hpa   2%/50%    1         10        1          4m3s

5：進行壓測查看pod的數量變化

1：查看svc地址為：10.100.114.140
2：進行壓力測試：
while true; do wget -q -O- http://10.100.114.140:8080/test/index.html; done

3：查看hpa負載
-6-95 ~]# kubectl get hpa -n app
NAME         REFERENCE               TARGETS    MINPODS   MAXPODS   REPLICAS   AGE
tomcat-hpa   Deployment/tomcat-hpa   210%/50%   1         10        1          4m8s

4：查看pod數量變化
-6-95 ~]# kubectl get pods -n app -o wide
NAME                          READY   STATUS    RESTARTS   AGE     IP           NODE                NOMINATED NODE   READINESS GATES
tomcat-hpa-664cc9d9f9-24r8s   1/1     Running   0          6s      172.6.95.8   k8s-6-95.host.com   <none>           <none>
tomcat-hpa-664cc9d9f9-4xpsb   1/1     Running   0          6s      172.6.95.7   k8s-6-95.host.com   <none>           <none>
tomcat-hpa-664cc9d9f9-8w7tk   1/1     Running   0          7m52s   172.6.94.5   k8s-6-94.host.com   <none>           <none>
tomcat-hpa-664cc9d9f9-gk9k6   1/1     Running   0          6s      172.6.94.6   k8s-6-94.host.com   <none>           <none>

5：測試停止后，pod數量減少為1
-6-95 ~]# kubectl get pods -n app -o wide
NAME                          READY   STATUS    RESTARTS   AGE     IP           NODE                NOMINATED NODE   READINESS GATES
tomcat-hpa-664cc9d9f9-4xpsb   1/1     Running   0          8m36s   172.6.95.7   k8s-6-95.host.com   <none>           <none>

注釋：如果要刪除存在的HPA：kubectl delete hpa tomcat-hpa -n app

四：基於內存自動擴容

1：修改資源配置清單

-6-96 tomcat-hpa]# cat dp.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
  name: tomcat-hpa
  namespace: app
spec:
  selector:
    matchLabels:
      app: tomcat-hpa
  template:
    metadata:
      labels:
        app: tomcat-hpa
    spec:
      containers:
      - name: tomcat-hpa
        image: harbor.auth.com/base/tomcat:v1.0
        imagePullPolicy: IfNotPresent
        ports:
        - containerPort: 8080
        resources:
          limits:
            memory: 500Mi
          requests:
            memory: 300Mi

-6-96 tomcat-hpa]# cat tomcat-memory-hpa.yaml 
apiVersion: autoscaling/v2beta1
kind: HorizontalPodAutoscaler
metadata:
  name: tomcat-hpa
  namespace: app
spec:
  maxReplicas: 10
  minReplicas: 1
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: tomcat-hpa
  metrics:
  - type: Resource
    resource:
      name: memory
      targetAverageUtilization: 20

注釋：添加hpa資源配置清單。targetAverageUtilization: 20 表示使用內存的20%(也就是300Mi的20%)

2：應用資源配置清單

-6-94 ~]# kubectl apply -f http://k8s-yaml.auth.com/tomcat-hpa/dp.yaml
-6-94 ~]# kubectl apply -f http://k8s-yaml.auth.com/tomcat-hpa/tomcat-memory-hpa.yaml

3：驗證HPA，進行測試，觀察pod數量變化

-6-94 ~]# kubectl get hpa -n app
NAME         REFERENCE               TARGETS   MINPODS   MAXPODS   REPLICAS   AGE
tomcat-hpa   Deployment/tomcat-hpa   27%/20%   1         10        1          18s

-6-94 ~]# kubectl get pods -n app -o wide
NAME                          READY   STATUS    RESTARTS   AGE   IP           NODE                NOMINATED NODE   READINESS GATES
tomcat-hpa-59764c648b-2mqq4   1/1     Running   0          17s   172.6.95.7   k8s-6-95.host.com   <none>           <none>
tomcat-hpa-59764c648b-d4rxt   1/1     Running   0          17m   172.6.94.5   k8s-6-94.host.com   <none>           <none>