HPA Pod自动扩缩容

一、简介

我们可以通过手工执行 kubectl scale命令和在Dashboard 上操作实现pod的扩缩容，Kubernetes系统也能够根据Pod当前负载的变化情况自动的进行扩缩容。Kubernetes提供了一个资源对象：Horizontal Pod Autoscaling （Pod水平自动伸缩），简称 HPA。HPA通过监控分析 RC或者 Deployment控制的所有 Pod的负载变化情况确定是否需要调整 Pod 的副本数量，这是HPA的基本原理。

HPA 控制器通过 Metrics Server的API获取指标数据，基于用户定义的扩缩容规则进行计算，得到目标Pod的副本数量。当目标Pod副本数量与当前副本数量不同时，HPA 控制器就向Pod副本控制器（Deployment、RC或RS）发起scale操作，调整Pod的副本数量，完成扩缩容操作。

二、部署操作

2.1 hpa-nginx.yaml

---
apiVersion: v1
kind: Service
metadata:
  name: hpa-nginx
  labels:
    app: hpa-nginx
spec:
  selector:
    app: hpa-nginx
  ports:
  - port: 80
    protocol: TCP
    targetPort: 80

---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: hpa-nginx
  labels:
    app: hpa-nginx
spec:
  selector:
    matchLabels:
      app: hpa-nginx
  revisionHistoryLimit: 10
  template:
    metadata:
      name: hpa-nginx
      labels:
        app: hpa-nginx
    spec:
      containers:
      - name: hpa-nginx
        image:  nginx:1.18.0
        ports:
        - containerPort: 80
        resources:
          limits:
            cpu: 0.2
            memory: 100Mi
          requests:
            cpu: 0.1
            memory: 10Mi
            
# resources 字段需要添加，否则hpa无法获取目标pod的资源利用指标，会显示unknow


# kubectl get pods -l app=hpa-nginx
NAME                         READY   STATUS    RESTARTS   AGE
hpa-nginx-5dc8fcf8c8-wl9jr   1/1     Running   0          10m


# kubectl get svc -l app=hpa-nginx
NAME        TYPE        CLUSTER-IP      EXTERNAL-IP   PORT(S)   AGE
hpa-nginx   ClusterIP   10.100.92.206   <none>        80/TCP    11m

2.2 创建一个HPA

# kubectl autoscale deployment hpa-nginx --cpu-percent=8 --min=1  --max=10


# kubectl get hpa
NAME        REFERENCE              TARGETS   MINPODS   MAXPODS   REPLICAS   AGE
hpa-nginx   Deployment/hpa-nginx   0%/10%    1         10        1          13h

此命令创建一个关联资源hpa-nginx 的 HPA，最小的Pod副本数量为1，最大为10。HPA会根据设定的cpu使用率（8%）动态的增加或减少Pod的数量。

2.3 测试HPA自动扩缩容

hpa-verify.yaml

apiVersion: v1
kind: Pod
metadata:
  name: hpa-busybox
  labels:
    app: hpa-busybox
spec:
  containers:
  - name: hpa-busybox
    image: registry.cn-hangzhou.aliyuncs.com/google_containers/busybox
    imagePullPolicy: IfNotPresent
    command:
    - sleep
    - "600"

kubectl exec -it hpa-busybox-d9c949d7-mpplg -- sh
/ # while true; do wget -q -O- http://hpa-nginx;done

查看Pod资源使用情况和hpa扩缩容情况

# hpa情况
# kubectl get hpa
NAME        REFERENCE              TARGETS   MINPODS   MAXPODS   REPLICAS   AGE
hpa-nginx   Deployment/hpa-nginx   86%/10%   1         10        9          14h

# Pod资源情况

# kubectl top pod hpa-nginx-57fcdcdc46-ww5s6
NAME                         CPU(cores)   MEMORY(bytes)   
hpa-nginx-57fcdcdc46-ww5s6   86m          1Mi 

# hpa扩缩容情况
# kubectl get pod -l app=hpa-nginx
NAME                         READY   STATUS    RESTARTS   AGE
hpa-nginx-57fcdcdc46-5frjt   1/1     Running   0          43s
hpa-nginx-57fcdcdc46-856lm   1/1     Running   0          28s
hpa-nginx-57fcdcdc46-9fm5x   1/1     Running   0          28s
hpa-nginx-57fcdcdc46-bq48j   1/1     Running   0          43s
hpa-nginx-57fcdcdc46-gx48s   1/1     Running   0          28s
hpa-nginx-57fcdcdc46-h7kv4   1/1     Running   0          28s
hpa-nginx-57fcdcdc46-pthrz   1/1     Running   0          13s
hpa-nginx-57fcdcdc46-rwz7k   1/1     Running   0          43s
hpa-nginx-57fcdcdc46-ww5s6   1/1     Running   0          14h

将测试访问停掉之后，过段时间就会自动缩容

# kubectl get hpa
NAME        REFERENCE              TARGETS   MINPODS   MAXPODS   REPLICAS   AGE
hpa-nginx   Deployment/hpa-nginx   0%/10%    1         10        1          14h

定义申明式yaml文件来创建hpa

apiVersion: autoscaling/v1
kind: HorizontalPodAutoscaler
metadata:
  name: hpa-nginx
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: hpa-nginx
  minReplicas: 1
  maxReplicas: 10
  targetCPUUtilizationPercentage: 50