当前位置： 首页 > news >正文

30Metrics Server的使用

news 来源：原创 2025/5/10 4:49:33

一、准备工作

1、下载Metrics Server镜像

[root@master 30]# cat metrics-server.sh
repo=registry.aliyuncs.com/google_containersname=k8s.gcr.io/metrics-server/metrics-server:v0.6.1
src_name=metrics-server:v0.6.1docker pull $repo/$src_namedocker tag $repo/$src_name $name
docker rmi $repo/$src_name
[root@master 30]# ./metrics-server.sh
v0.6.1: Pulling from google_containers/metrics-server
2df365faf0e3: Already exists
0ae7e0717edb: Pull complete
Digest: sha256:5ddc6458eb95f5c70bd13fdab90cbd7d6ad1066e5b528ad1dcb28b76c5fb2f00
Status: Downloaded newer image for registry.aliyuncs.com/google_containers/metrics-server:v0.6.1
registry.aliyuncs.com/google_containers/metrics-server:v0.6.1
Untagged: registry.aliyuncs.com/google_containers/metrics-server:v0.6.1
Untagged: registry.aliyuncs.com/google_containers/metrics-server@sha256:5ddc6458eb95f5c70bd13fdab90cbd7d6ad1066e5b528ad1dcb28b76c5fb2f00

2、下载httpd:alpine镜像（hpa会用到这个镜像）

[root@master 30]# docker pull httpd:alpine
alpine: Pulling from library/httpd
f18232174bc9: Pull complete
f6f8d7d49e24: Pull complete
aca02fb0fe83: Pull complete
4f4fb700ef54: Pull complete
da9bd0c8aef2: Pull complete
e7c0ad6e3e09: Pull complete
852b1f8ff649: Pull complete
Digest: sha256:4aec2953509e2d3aa5a8d73c580a381be44803fd2481875b15d9ad7d2810d7ca
Status: Downloaded newer image for httpd:alpine
docker.io/library/httpd:alpine

3、下载并修改YAML文件

  • 浏览器进入git地址：
    https://github.com/kubernetes-sigs/metrics-server/releases/latest/
  • 下载components.yaml文件
  • 修改Metrics Server的通信协议，追加参数–kubelet-insecure-tls
 spec:template:spec:containers:- args:- --kubelet-insecure-tls
  • 修改镜像地址为阿里云的镜像地址
image: registry.aliyuncs.com/google_containers/metrics-server:v0.6.1

二、部署Metrics Server

[root@master 30]# kubectl apply -f components.yaml
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
[root@master 30]# kubectl get pod -n kube-system | grep metrics-server
metrics-server-66bdd9d576-d6rff   0/1     Running   0               30s

三、使用Metrics Server

1、查看节点的资源使用率

[root@master 30]# kubectl top node
NAME     CPU(cores)   CPU%   MEMORY(bytes)   MEMORY%
master   185m         9%     1242Mi          33%
worker   62m          3%     656Mi           38%

2、查看Pod的资源使用率

[root@master 30]# kubectl top pod -n kube-system
NAME                              CPU(cores)   MEMORY(bytes)
coredns-64897985d-4dx65           3m           16Mi
coredns-64897985d-89vss           3m           17Mi
etcd-master                       19m          80Mi
kube-apiserver-master             54m          305Mi
kube-controller-manager-master    30m          53Mi
kube-proxy-8ght4                  1m           21Mi
kube-proxy-g8mlb                  1m           17Mi
kube-scheduler-master             4m           22Mi
metrics-server-66bdd9d576-d6rff   5m           14Mi

四、HorizontalPodAutoscaler（hpa）

专门用来自动伸缩Pod的数量，适用于Deployment和StatefulSet，但不能用于DaemonSet。

HorizontalPodAutoscaler的能力完全基于Metrics Server，它从Metrics Server获取当前应用的运行指标，主要是CPU使用率，再依据预定的策略增加或者减少Pod的数量。

1、创建一个Nginx应用，作为自动伸缩的目标对象

[root@master 30]# cat 03ngx-hpa-dep.yml
apiVersion: apps/v1
kind: Deployment
metadata:name: ngx-hpa-dep
spec:replicas: 1selector:matchLabels:app: ngx-hpa-deptemplate:metadata:labels:app: ngx-hpa-depspec:containers:- image: nginx:alpinename: nginxports:- containerPort: 80resources:requests:cpu: 50mmemory: 10Milimits:cpu: 100mmemory: 20Mi
---apiVersion: v1
kind: Service
metadata:name: ngx-hpa-svc
spec:ports:- port: 80protocol: TCPtargetPort: 80selector:app: ngx-hpa-dep
[root@master 30]# kubectl apply -f 03ngx-hpa-dep.yml
deployment.apps/ngx-hpa-dep created
service/ngx-hpa-svc created
[root@master 30]# kubectl get svc
NAME          TYPE        CLUSTER-IP       EXTERNAL-IP   PORT(S)        AGE
kubernetes    ClusterIP   10.96.0.1        <none>        443/TCP        7d7h
ngx-hpa-svc   ClusterIP   10.108.185.201   <none>        80/TCP         11s
ngx-svc       NodePort    10.102.24.158    <none>        80:30749/TCP   163m
[root@master 30]# kubectl get pod
NAME                           READY   STATUS    RESTARTS   AGE
ngx-dep-6796688696-2xjz8       1/1     Running   0          167m
ngx-dep-6796688696-pkwft       1/1     Running   0          158m
ngx-hpa-dep-86f66c75f5-kwgln   1/1     Running   0          18s
redis-ds-g8mt5                 1/1     Running   0          3h5m
redis-ds-ztt5f                 1/1     Running   0          3h5m

2、生成YAML样板文件

  • min，Pod数量的最小值，也就是缩容的下限。
  • max，Pod数量的最大值，也就是扩容的上限。
  • cpu-percent，CPU使用率指标，当大于这个值时扩容，小于这个值时缩容。
[root@master 30]# kubectl autoscale deploy ngx-hpa-dep --min=2 --max=10 --cpu-percent=5 --dry-run=client -o yaml
apiVersion: autoscaling/v1
kind: HorizontalPodAutoscaler
metadata:creationTimestamp: nullname: ngx-hpa-dep
spec:maxReplicas: 10minReplicas: 2scaleTargetRef:apiVersion: apps/v1kind: Deploymentname: ngx-hpa-deptargetCPUUtilizationPercentage: 5
status:currentReplicas: 0desiredReplicas: 0

3、创建HorizontalPodAutoscale

[root@master 30]# cat 04ngx-hpa.yml
apiVersion: autoscaling/v1
kind: HorizontalPodAutoscaler
metadata:name: ngx-hpa
spec:maxReplicas: 10minReplicas: 2scaleTargetRef:apiVersion: apps/v1kind: Deploymentname: ngx-hpa-deptargetCPUUtilizationPercentage: 5
[root@master 30]# kubectl apply -f 04ngx-hpa.yml
horizontalpodautoscaler.autoscaling/ngx-hpa created
[root@master 30]# kubectl get deploy ngx-hpa-dep
NAME          READY   UP-TO-DATE   AVAILABLE   AGE
ngx-hpa-dep   2/2     2            2           12m
[root@master 30]# kubectl get hpa ngx-hpa
NAME      REFERENCE                TARGETS   MINPODS   MAXPODS   REPLICAS   AGE
ngx-hpa   Deployment/ngx-hpa-dep   0%/5%     2         10        2          68s

下面我们来给Nginx加上压力流量，运行一个测试Pod，使用的镜像是“httpd:alpine”，它里面有HTTP性能测试工具ab（Apache Bench）：

[root@master ~]# kubectl run test1 -it --image=httpd:alpine -- sh
If you don't see a command prompt, try pressing enter.
/usr/local/apache2 # ab -V
This is ApacheBench, Version 2.3 <$Revision: 1923142 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org//usr/local/apache2 #

然后我们向Nginx发送一百万个请求，持续1分钟，再用kubectl get hpa来观察HorizontalPodAutoscaler的运行状况：

/usr/local/apache2 # ab -c 10 -t 60 -n 1000000 'http://ngx-hpa-svc/'
This is ApacheBench, Version 2.3 <$Revision: 1923142 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/Benchmarking ngx-hpa-svc (be patient)
Completed 100000 requests
Completed 200000 requests
Completed 300000 requests
Completed 400000 requests
Finished 466952 requestsServer Software:        nginx/1.27.3
Server Hostname:        ngx-hpa-svc
Server Port:            80Document Path:          /
Document Length:        615 bytesConcurrency Level:      10
Time taken for tests:   60.000 seconds
Complete requests:      466952
Failed requests:        6(Connect: 0, Receive: 0, Length: 0, Exceptions: 6)
Total transferred:      395981232 bytes
HTML transferred:       287179785 bytes
Requests per second:    7782.53 [#/sec] (mean)
Time per request:       1.285 [ms] (mean)
Time per request:       0.128 [ms] (mean, across all concurrent requests)
Transfer rate:          6445.00 [Kbytes/sec] receivedConnection Times (ms)min  mean[+/-sd] median   max
Connect:        0    0   0.3      0      21
Processing:     0    1   6.2      0      88
Waiting:        0    1   6.1      0      88
Total:          0    1   6.2      1      88Percentage of the requests served within a certain time (ms)50%      166%      175%      180%      190%      195%      298%      399%     12100%     88 (longest request)

[root@master 20]# kubectl get hpa ngx-hpa -w
NAME      REFERENCE                TARGETS   MINPODS   MAXPODS   REPLICAS   AGE
ngx-hpa   Deployment/ngx-hpa-dep   0%/5%     2         10        2          32m
ngx-hpa   Deployment/ngx-hpa-dep   19%/5%    2         10        2          32m
ngx-hpa   Deployment/ngx-hpa-dep   178%/5%   2         10        4          32m
ngx-hpa   Deployment/ngx-hpa-dep   195%/5%   2         10        8          33m
ngx-hpa   Deployment/ngx-hpa-dep   188%/5%   2         10        10         33m
ngx-hpa   Deployment/ngx-hpa-dep   144%/5%   2         10        10         33m
ngx-hpa   Deployment/ngx-hpa-dep   35%/5%    2         10        10         33m
ngx-hpa   Deployment/ngx-hpa-dep   3%/5%     2         10        10         34m
ngx-hpa   Deployment/ngx-hpa-dep   0%/5%     2         10        10         34m
ngx-hpa   Deployment/ngx-hpa-dep   0%/5%     2         10        10         38m
ngx-hpa   Deployment/ngx-hpa-dep   0%/5%     2         10        6          39m
ngx-hpa   Deployment/ngx-hpa-dep   0%/5%     2         10        2          39m

因为Metrics Server大约每15秒采集一次数据，所以HorizontalPodAutoscaler的自动化扩容和缩容也是按照这个时间点来逐步处理的。

当它发现目标的CPU使用率超过了预定的5%后，就会以2的倍数开始扩容，一直到数量上限，然后持续监控一段时间，如果CPU使用率回落，就会再缩容到最小值。

相关文章：

  • 在VirtualBox上安装Ubuntu
  • Electron 中引入MessageChannel 大大缩短不同渲染进程和 Webview 各组件 1o1的通信链路
  • 山东大学软件学院创新项目实训开发日志（18）之对话自动生成标题设为用户第一次对话发的文字
  • C++11：模板元编程（TMP）基础
  • 深入理解C++数组：从基础到实践
  • 【已更新】2025华中杯C题数学建模网络挑战赛思路代码文章教学数学建模思路:就业状态分析与预测
  • Breeze 55A FOC 电调：无人机动力控制的高效核心方案
  • 蓝桥杯题目：二维前缀和
  • 【PyQt5】QLineEdit文本对话框点击时关联槽函数，槽函数打开文件选择对话框；并解决选择文件后闪退的问题
  • 【杂谈】-自动驾驶变革：货运革新与机器人出租车崛起
  • 非洲电商争夺战：中国闪电战遭遇本土游击队的降维打击
  • 合成数据在自动驾驶中的实践：工作流、关键技术与评估体系全解析
  • 四、小白如何用Pygame制作一款跑酷类游戏（页面暂停和主角跑步动作的实现）
  • 性能测试中TPS、并发数与线程数的关系
  • 状态模式：有限状态机在电商订单系统中的设计与实现
  • 树莓派超全系列教程文档--(29)config.txt介绍
  • C/C++指针
  • Spring AI与通义千问的完美结合：构建智能对话应用
  • 斯托克斯矢量只定义在线极化基下
  • 美信监控易：运维管理软件的售后优势
  • 比特币价格重返10万美元，哪些因素使然？
  • 本周看啥｜喜欢二次元的观众，去电影院吧
  • 安徽亳州涡阳县司法局党组书记刘兴连落马
  • 欧盟公布对美关税反制清单，瞄准美国飞机等产品
  • 现场丨“影像上海”启幕：串联摄影、电影与当代艺术
  • 《2025城市青年旅行消费报告》发布，解码青年出行特征