资源调度-HPA：基于负载指标自动水平扩容缩容Pod

Pod 自动扩容：可以根据 CPU 使用率或自定义指标（metrics）自动对 Pod 进行扩/缩容。
控制管理器每隔30s（可以通过–horizontal-pod-autoscaler-sync-period修改）查询metrics的资源使用情况

• 支持三种metrics类型
  • 预定义metrics（比如Pod的CPU）以利用率的方式计算
  • 自定义的Pod metrics，以原始值（raw value）的方式计算
  • 自定义的object metrics
• 支持两种metrics查询方式：Heapster和自定义的REST API
• 支持多metrics

通常用于 Deployment，不适用于无法扩/缩容的对象，如 DaemonSet

实现 cpu 或内存的监控，首先有个前提条件是该对象必须配置了 resources.requests.cpu 或 resources.requests.memory 才可以，可以配置当 cpu/memory 达到上述配置的百分比后进行扩容或缩容

1.创建一个 HPA：

#修改配置文件 #加入资源限制
[root@k8s-master deployments]# vi nginx-deploy.yaml

apiVersion: apps/v1
kind: Deployment
metadata:
  labels:
    app: nginx-deploy
  name: nginx-deploy
  namespace: default
spec:
  replicas: 1
  revisionHistoryLimit: 10
  selector:
    matchLabels:
      app: nginx-deploy
  strategy:
    rollingUpdate:
      maxSurge: 25%
      maxUnavailable: 25%
    type: RollingUpdate
  template:
    metadata:
      labels:
        app: nginx-deploy
    spec:
      containers:
      - image: nginx:1.7.9
        imagePullPolicy: IfNotPresent
        name: nginx
        #加入资源限制
        resources:
          limits:
            cpu: 200m
            memory: 128Mi
          requests:
            cpu: 10m
            memory: 128Mi   
      restartPolicy: Always
      terminationGracePeriodSeconds: 30

# 更新操作
[root@k8s-master deployments]# kubectl replace -f nginx-deploy.yaml
deployment.apps/nginx-deploy replaced

# 查看 只有一个副本
[root@k8s-master deployments]# kubectl get deploy nginx-deploy
NAME           READY   UP-TO-DATE   AVAILABLE   AGE
nginx-deploy   1/1     1            1           18h

创建一个 HPA：

1. 先准备一个好一个有做资源限制的 deployment
2. 执行命令 kubectl autoscale deploy nginx-deploy --cpu-percent=20 --min=2 --max=5
3. 通过 kubectl get hpa 可以获取 HPA 信息

[root@k8s-master deployments]#  kubectl autoscale deploy nginx-deploy --cpu-percent=20 --min=2 --max=5
horizontalpodautoscaler.autoscaling/nginx-deploy autoscaled
#现在再次查看 副本自动扩容到了2个
[root@k8s-master deployments]#  kubectl get deploy nginx-deploy
NAME           READY   UP-TO-DATE   AVAILABLE   AGE
nginx-deploy   2/2     2            2           18h

未安装 Metrics Server 查看 pod 内存、CPU 将报错

[root@k8s ~]# kubectl top pods --all-namespaces
error: Metrics API not available

Pod 的水平自动缩放（前提需要安装插件metrics-server ）

2.Metrics Server安装

一、使用阿里云的components.yaml配置文件

aliyun-components.yaml

[root@k8s-master ~]# rz
[root@k8s-master ~]# ls
aliyun-components.yaml  anaconda-ks.cfg
#查看文件
[root@k8s-master ~]# cat aliyun-components.yaml

二、执行安装命令

[root@k8s-master ~]# kubectl apply -f aliyun-components.yaml 

执行完输出内容:

serviceaccount/metrics-server created
clusterrole.rbac.authorization.k8s.io/system:aggregated-metrics-reader created
clusterrole.rbac.authorization.k8s.io/system:metrics-server created
rolebinding.rbac.authorization.k8s.io/metrics-server-auth-reader created
clusterrolebinding.rbac.authorization.k8s.io/metrics-server:system:auth-delegator created
clusterrolebinding.rbac.authorization.k8s.io/system:metrics-server created
service/metrics-server created
deployment.apps/metrics-server created
apiservice.apiregistration.k8s.io/v1beta1.metrics.k8s.io created

三、查看效果，验证是否安装成功

[root@k8s-master ~]# kubectl get svc metrics-server -n kube-system
NAME             TYPE        CLUSTER-IP     EXTERNAL-IP   PORT(S)   AGE
metrics-server   ClusterIP   10.105.89.80   <none>        443/TCP   152m
# 使用kubectl top node 查看资源使用情况
[root@k8s-master ~]# kubectl top node
NAME         CPU(cores)   CPU%   MEMORY(bytes)   MEMORY%   
k8s-master   91m          4%     912Mi           24%       
k8s-node1    31m          1%     756Mi           20%       
[root@k8s-master ~]# kubectl top pods --all-namespaces
NAMESPACE      NAME                                 CPU(cores)   MEMORY(bytes)   
kube-flannel   kube-flannel-ds-8fxmx                3m           18Mi            
kube-flannel   kube-flannel-ds-xfmpr                4m           20Mi            
kube-system    coredns-65c54cc984-7vd4t             1m           17Mi            
kube-system    coredns-65c54cc984-qqv8m             1m           16Mi            
kube-system    etcd-k8s-master                      9m           67Mi            
kube-system    kube-apiserver-k8s-master            34m          228Mi           
kube-system    kube-controller-manager-k8s-master   12m          49Mi            
kube-system    kube-proxy-l99bw                     1m           17Mi            
kube-system    kube-proxy-x8tdf                     1m           16Mi            
kube-system    kube-scheduler-k8s-master            2m           20Mi            
kube-system    metrics-server-b9f7b695f-bljhg       3m           20Mi     

至此,metrics-server安装完成

[root@k8s-master deployments]# kubectl top pods
NAME                           CPU(cores)   MEMORY(bytes)               
nginx-deploy-56696fbb5-4ksdj   0m           1Mi             
nginx-deploy-56696fbb5-wtjtq   0m           1Mi                

3.创建一个servie

[root@k8s-master deployments]# kubectl create -f nginx-svc.yaml
service/nginx-svc created

apiVersion: v1
kind: Service
metadata:
  name: nginx-svc
  labels: 
    app: nginx
spec:
  selector:
    app: nginx-deploy
  ports:
  - port: 80
    targetPort: 80
    name: web
  type: NodePort

[root@k8s-master deployments]# kubectl get svc 
NAME            TYPE        CLUSTER-IP       EXTERNAL-IP   PORT(S)        AGE
nginx-svc       NodePort    10.100.220.235   <none>        80:31590/TCP   55s

[root@k8s-master deployments]# kubectl get pods -o wide
NAME                           READY   STATUS    RESTARTS   AGE     IP            NODE        NOMINATED NODE   READINESS GATES
nginx-deploy-56696fbb5-4ksdj   1/1     Running   0          37m     10.244.1.46   k8s-node1   <none>           <none>
nginx-deploy-56696fbb5-wtjtq   1/1     Running   0          40m     10.244.1.45   k8s-node1   <none>           <none>

4.测试

找到对应服务的 service，编写循环测试脚本提升内存与 cpu 负载

[root@k8s-node1 ~]# while true; do wget -q -O- http://10.100.220.235  > /dev/null ; done

[root@k8s-master deployments]# kubectl top pods
NAME                           CPU(cores)   MEMORY(bytes)           
nginx-deploy-56696fbb5-4ksdj   0m           1Mi             
nginx-deploy-56696fbb5-bdkx2   2m           1Mi             
nginx-deploy-56696fbb5-wtjtq   0m           1Mi             
nginx-deploy-56696fbb5-zrdb7   0m           1Mi             

#400%/20%
[root@k8s-master deployments]# kubectl get hpa
NAME           REFERENCE                 TARGETS    MINPODS   MAXPODS   REPLICAS   AGE
nginx-deploy   Deployment/nginx-deploy   400%/20%   2         5         2          67m

#查看已经扩容到5个副本了
[root@k8s-master deployments]# kubectl get deploy
NAME           READY   UP-TO-DATE   AVAILABLE   AGE
nginx-deploy   5/5     5            5           19h



#暂停之前死循环的命令 等一会再次查看  
[root@k8s-master deployments]# kubectl get hpa
NAME           REFERENCE                 TARGETS   MINPODS   MAXPODS   REPLICAS   AGE
nginx-deploy   Deployment/nginx-deploy   0%/20%    2         5         5          69m
#pod变化  
[root@k8s-master deployments]# kubectl get pods
NAME                          READY   STATUS    RESTARTS   AGE
nginx-deploy-c4986b7f-8tplt   1/1     Running   0          6m24s
nginx-deploy-c4986b7f-qltv6   1/1     Running   0          7m2s
#deployment变化
[root@k8s-master deployments]#  kubectl get deploy nginx-deploy
NAME           READY   UP-TO-DATE   AVAILABLE   AGE
nginx-deploy   2/2     2            2           19h

#发现已经自动完整了扩容缩容