3、Kubernetes 集群部署 Prometheus 和 Grafana

实验环境
控制节点/master01 192.168.110.10
工作节点/node01 192.168.110.20
工作节点/node02 192.168.110.30

node-exporter 安装

#创建监控 namespace
kubectl create ns monitor-sa#部署 node-exporter
mkdir /opt/prometheus
cd /opt/prometheus/vim node-export.yaml
---
apiVersion: apps/v1
kind: DaemonSet				#可以保证 k8s 集群的每个节点都运行完全一样的 pod
metadata:name: node-exporternamespace: monitor-salabels:name: node-exporter
spec:selector:matchLabels:name: node-exportertemplate:metadata:labels:name: node-exporterspec:hostPID: truehostIPC: truehostNetwork: truecontainers:- name: node-exporterimage: prom/node-exporter:v0.16.0ports:- containerPort: 9100resources:requests:cpu: 0.15		#这个容器运行至少需要0.15核cpusecurityContext:privileged: true	#开启特权模式args:- --path.procfs- /host/proc- --path.sysfs- /host/sys- --collector.filesystem.ignored-mount-points- '"^/(sys|proc|dev|host|etc)($|/)"'volumeMounts:- name: devmountPath: /host/dev- name: procmountPath: /host/proc- name: sysmountPath: /host/sys- name: rootfsmountPath: /rootfstolerations:- key: "node-role.kubernetes.io/master"operator: "Exists"effect: "NoSchedule"volumes:- name: prochostPath:path: /proc- name: devhostPath:path: /dev- name: syshostPath:path: /sys- name: rootfshostPath:path: /kubectl apply -f node-export.yamlkubectl get pods -n monitor-sa -o wide[root@master01 prometheus]# kubectl get pods -n monitor-sa -o wide
NAME                  READY   STATUS    RESTARTS   AGE     IP               NODE             NOMINATED NODE   READINESS GATES
node-exporter-4mjm8   1/1     Running   5          3m40s   192.168.110.30   192.168.110.30   <none>           <none>
node-exporter-kzbpv   1/1     Running   5          3m40s   192.168.110.20   192.168.110.20   <none>           <none>

通过 node-exporter 采集数据
node-exporter 默认的监听端口是 9100，可以执行 curl http://主机ip:9100/metrics 获取到主机的所有监控数据

curl -Ls http://192.168.110.20:9100/metrics | grep node_cpu_seconds
curl -Ls http://192.168.110.20:9100/metrics | grep node_load[root@node02 ~]# curl -Ls http://192.168.110.20:9100/metrics | grep node_cpu_seconds
# HELP node_cpu_seconds_total Seconds the cpus spent in each mode.
# TYPE node_cpu_seconds_total counter
node_cpu_seconds_total{cpu="0",mode="idle"} 67314.31
node_cpu_seconds_total{cpu="0",mode="iowait"} 5.25
... ...
node_cpu_seconds_total{cpu="3",mode="system"} 768.89
node_cpu_seconds_total{cpu="3",mode="user"} 143.63
[root@node02 ~]# curl -Ls http://192.168.110.20:9100/metrics | grep node_load
# HELP node_load1 1m load average.
# TYPE node_load1 gauge
node_load1 0.41
# HELP node_load15 15m load average.
# TYPE node_load15 gauge
node_load15 0.75
# HELP node_load5 5m load average.
# TYPE node_load5 gauge
node_load5 0.62

Prometheus 安装和配置

创建 sa 账号，对 sa 做 rbac 授权

#创建一个 sa 账号 monitor
kubectl create serviceaccount monitor -n monitor-sa
#把 sa 账号 monitor 通过 clusterrolebing 绑定到 clusterrole 上
kubectl create clusterrolebinding monitor-clusterrolebinding -n monitor-sa --clusterrole=cluster-admin  --serviceaccount=monitor-sa:monitor

创建一个 configmap 存储卷，用来存放 prometheus 配置信息

vim prometheus-cfg.yaml
---
kind: ConfigMap
apiVersion: v1
metadata:labels:app: prometheusname: prometheus-confignamespace: monitor-sa
data:prometheus.yml: |global:                     #指定prometheus的全局配置，比如采集间隔，抓取超时时间等scrape_interval: 15s      #采集目标主机监控数据的时间间隔，默认为1mscrape_timeout: 10s       #数据采集超时时间，默认10sevaluation_interval: 1m 	#触发告警生成alert的时间间隔，默认是1mscrape_configs:             #配置数据源，称为target，每个target用job_name命名。又分为静态配置和服务发现- job_name: 'kubernetes-node'kubernetes_sd_configs:    # *_sd_configs 指定的是k8s的服务发现- role: node              #使用node角色，它使用默认的kubelet提供的http端口来发现集群中每个node节点relabel_configs:	        #重新标记- source_labels: [__address__]    #配置的原始标签，匹配地址regex: '(.*):10250'             #匹配带有10250端口的urlreplacement: '${1}:9100'        #把匹配到的ip:10250的ip保留target_label: __address__       #新生成的url是${1}获取到的ip:9100action: replace         #动作替换- action: labelmapregex: __meta_kubernetes_node_label_(.+)    #匹配到下面正则表达式的标签会被保留,如果不做regex正则的话，默认只是会显示instance标签- job_name: 'kubernetes-node-cadvisor'    #抓取cAdvisor数据，是获取kubelet上/metrics/cadvisor接口数据来获取容器的资源使用情况kubernetes_sd_configs:- role:  nodescheme: httpstls_config:ca_file: /var/run/secrets/kubernetes.io/serviceaccount/ca.crtbearer_token_file: /var/run/secrets/kubernetes.io/serviceaccount/tokenrelabel_configs:- action: labelmap    #把匹配到的标签保留regex: __meta_kubernetes_node_label_(.+)    #保留匹配到的具有__meta_kubernetes_node_label的标签- target_label: __address__                   #获取到的地址：__address__="192.168.80.20:10250"replacement: kubernetes.default.svc:443     #把获取到的地址替换成新的地址kubernetes.default.svc:443- source_labels: [__meta_kubernetes_node_name]regex: (.+)                                 #把原始标签中__meta_kubernetes_node_name值匹配到target_label: __metrics_path__              #获取__metrics_path__对应的值replacement: /api/v1/nodes/${1}/proxy/metrics/cadvisor	#把metrics替换成新的值api/v1/nodes/<node_name>/proxy/metrics/cadvisor#${1}是__meta_kubernetes_node_name获取到的值#最后通过https://<apiserver_address>/api/v1/nodes/<node_name>/proxy/metrics/cadvisor来获取对应节点cadvisor监控数据- job_name: 'kubernetes-apiserver'kubernetes_sd_configs:- role: endpoints                             #使用k8s中的endpoint服务发现，采集apiserver 6443端口获取到的数据scheme: httpstls_config:ca_file: /var/run/secrets/kubernetes.io/serviceaccount/ca.crtbearer_token_file: /var/run/secrets/kubernetes.io/serviceaccount/tokenrelabel_configs:- source_labels: [__meta_kubernetes_namespace, __meta_kubernetes_service_name, __meta_kubernetes_endpoint_port_name]    #[endpoint这个对象的名称空间,endpoint对象的服务名,exnpoint的端口名称]action: keep    #采集满足条件的实例，其他实例不采集regex: default;kubernetes;https    #正则匹配到的默认空间下的service名字是kubernetes，协议是https的endpoint类型保留下来- job_name: 'kubernetes-service-endpoints'kubernetes_sd_configs:- role: endpointsrelabel_configs:- source_labels: [__meta_kubernetes_service_annotation_prometheus_io_scrape]action: keepregex: true#重新打标仅抓取到的具有"prometheus.io/scrape: true"的annotation的端点， 意思是说如果某个service具有prometheus.io/scrape = true的annotation声明则抓取，annotation本身也是键值结构， 所以这里的源标签设置为键，而regex设置值true，当值匹配到regex设定的内容时则执行keep动作也就是保留，其余则丢弃。- source_labels: [__meta_kubernetes_service_annotation_prometheus_io_scheme]action: replacetarget_label: __scheme__regex: (https?)#重新设置scheme，匹配源标签__meta_kubernetes_service_annotation_prometheus_io_scheme也就是prometheus.io/scheme annotation，如果源标签的值匹配到regex，则把值替换为__scheme__对应的值。- source_labels: [__meta_kubernetes_service_annotation_prometheus_io_path]action: replacetarget_label: __metrics_path__regex: (.+)#应用中自定义暴露的指标，也许你暴露的API接口不是/metrics这个路径，那么你可以在这个POD对应的service中做一个 "prometheus.io/path = /mymetrics" 声明，上面的意思就是把你声明的这个路径赋值给__metrics_path__， 其实就是让prometheus来获取自定义应用暴露的metrices的具体路径， 不过这里写的要和service中做好约定，如果service中这样写 prometheus.io/app-metrics-path: '/metrics' 那么你这里就要__meta_kubernetes_service_annotation_prometheus_io_app_metrics_path这样写。- source_labels: [__address__, __meta_kubernetes_service_annotation_prometheus_io_port]action: replacetarget_label: __address__regex: ([^:]+)(?::\d+)?;(\d+)replacement: $1:$2#暴露自定义的应用的端口，就是把地址和你在service中定义的 "prometheus.io/port = <port>" 声明做一个拼接， 然后赋值给__address__，这样prometheus就能获取自定义应用的端口，然后通过这个端口再结合__metrics_path__来获取指标，如果__metrics_path__值不是默认的/metrics那么就要使用上面的标签替换来获取真正暴露的具体路径。- action: labelmap        #保留下面匹配到的标签regex: __meta_kubernetes_service_label_(.+)- source_labels: [__meta_kubernetes_namespace]action: replace        #替换__meta_kubernetes_namespace变成kubernetes_namespacetarget_label: kubernetes_namespace- source_labels: [__meta_kubernetes_service_name]action: replacetarget_label: kubernetes_namekubectl apply -f prometheus-cfg.yaml

通过 deployment 部署 prometheus

#将 prometheus 调度到 node1 节点，在 node1 节点创建 prometheus 数据存储目录
mkdir /data && chmod 777 /data#通过 deployment 部署 prometheus
vim prometheus-deploy.yaml
---
apiVersion: apps/v1
kind: Deployment
metadata:name: prometheus-servernamespace: monitor-salabels:app: prometheus
spec:replicas: 1selector:matchLabels:app: prometheuscomponent: server#matchExpressions:#- {key: app, operator: In, values: [prometheus]}#- {key: component, operator: In, values: [server]}template:metadata:labels:app: prometheuscomponent: serverannotations:prometheus.io/scrape: 'false'spec:nodeName: 192.168.110.20              #指定pod调度到哪个节点上	serviceAccountName: monitorcontainers:- name: prometheusimage: prom/prometheus:v2.2.1imagePullPolicy: IfNotPresentcommand:- prometheus- --config.file=/etc/prometheus/prometheus.yml- --storage.tsdb.path=/prometheus        #数据存储目录- --storage.tsdb.retention=720h          #数据保存时长- --web.enable-lifecycle                 #开启热加载ports:- containerPort: 9090protocol: TCPvolumeMounts:- mountPath: /etc/prometheus/prometheus.ymlname: prometheus-configsubPath: prometheus.yml- mountPath: /prometheus/name: prometheus-storage-volumevolumes:- name: prometheus-configconfigMap:name: prometheus-configitems:- key: prometheus.ymlpath: prometheus.ymlmode: 0644- name: prometheus-storage-volumehostPath:path: /datatype: Directorykubectl apply -f prometheus-deploy.yamlkubectl get pods -o wide -n monitor-sa [root@master01 prometheus]# kubectl get pods -o wide -n monitor-sa
NAME                                 READY   STATUS    RESTARTS   AGE   IP               NODE             NOMINATED NODE   READINESS GATES
node-exporter-4mjm8                  1/1     Running   5          15m   192.168.110.30   192.168.110.30   <none>           <none>
node-exporter-kzbpv                  1/1     Running   5          15m   192.168.110.20   192.168.110.20   <none>           <none>
prometheus-server-75fb7f8fc6-9rrl7   0/1     Pending   0          49s   <none>           node01           <none>           <none>

给 prometheus pod 创建一个 service

vim prometheus-svc.yaml
---
apiVersion: v1
kind: Service
metadata:name: prometheusnamespace: monitor-salabels:app: prometheus
spec:type: NodePortports:- port: 9090targetPort: 9090protocol: TCPnodePort: 31000selector:app: prometheuscomponent: serverkubectl apply -f prometheus-svc.yamlkubectl get pods -o wide -n monitor-sa 
kubectl get svc -n monitor-sa[root@master01 prometheus]# kubectl get pods -o wide -n monitor-sa
NAME                                 READY   STATUS    RESTARTS   AGE     IP               NODE             NOMINATED NODE   READINESS GATES
node-exporter-7lfl8                  1/1     Running   9          2d12h   192.168.110.20   192.168.110.20   <none>           <none>
node-exporter-hnpm2                  1/1     Running   0          2d12h   192.168.110.30   192.168.110.30   <none>           <none>
prometheus-server-84b77ffcf8-mj5rp   1/1     Running   0          5m22s   10.244.60.102    192.168.110.20   <none>           <none>
[root@master01 prometheus]# kubectl get svc -n monitor-sa
NAME         TYPE       CLUSTER-IP   EXTERNAL-IP   PORT(S)          AGE
prometheus   NodePort   10.0.0.113   <none>        9090:31000/TCP   5d5h

如发现镜像拉取失败导致无法建立 pod 检查docker 版本，要使用19.xx.x-22.xx.x 版

通过上面可以看到 service 在 node 节点上映射的端口是 31000，这样我们访问 k8s 集群的 node 节点的 ip:31000，就可以访问到 prometheus 的 web ui 界面了。
浏览器访问 http://192.168.110.20:31000
点击页面的Status->Targets，如看到所有 Target 状态都为 UP，说明我们配置的服务发现可以正常采集数据

查询 K8S 集群中一分钟之内每个 Pod 的 CPU 使用率
sum by (name)( rate(container_cpu_usage_seconds_total{image!=“”, name!=“”}[1m] ) )

Prometheus 配置热加载

为了每次修改配置文件可以热加载prometheus，也就是不停止prometheus就可以使配置生效。想要使配置生效可用如下热加载命令：

kubectl get pods -n monitor-sa -o wide -l app=prometheus[root@master01 prometheus]# kubectl get pods -n monitor-sa -o wide -l app=prometheus
NAME                                 READY   STATUS    RESTARTS   AGE   IP              NODE             NOMINATED NODE   READINESS GATES
prometheus-server-84b77ffcf8-mj5rp   1/1     Running   0          13m   10.244.60.102   192.168.110.20   <none>           <none>

想要使配置生效可用如下命令热加载

curl -X POST -Ls http://10.244.60.102:9090/-/reload 

查看 log

kubectl logs -n monitor-sa prometheus-server-84b77ffcf8-mj5rp | grep "Loading configuration file" [root@master01 prometheus]# kubectl logs -n monitor-sa prometheus-server-84b77ffcf8-mj5rp | grep "Loading configuration file"
level=info ts=2025-02-19T14:18:55.772194011Z caller=main.go:588 msg="Loading configuration file" filename=/etc/prometheus/prometheus.yml

一般热加载速度比较慢，可以暴力重启prometheus，如修改上面的 prometheus-cfg.yaml 文件之后，可用如下命令：
可执行先强制删除，然后再通过 apply 更新

kubectl delete -f prometheus-cfg.yaml
kubectl delete -f prometheus-deploy.yaml
kubectl apply -f prometheus-cfg.yaml
kubectl apply -f prometheus-deploy.yaml

Grafana 安装

部署

vim grafana.yaml
---
apiVersion: apps/v1
kind: Deployment
metadata:name: monitoring-grafananamespace: kube-system
spec:replicas: 1selector:matchLabels:task: monitoringk8s-app: grafanatemplate:metadata:labels:task: monitoringk8s-app: grafanaspec:containers:- name: grafanaimage: grafana/grafana:5.0.4ports:- containerPort: 3000protocol: TCPvolumeMounts:- mountPath: /etc/ssl/certsname: ca-certificatesreadOnly: true- mountPath: /varname: grafana-storageenv:- name: INFLUXDB_HOSTvalue: monitoring-influxdb- name: GF_SERVER_HTTP_PORTvalue: "3000"# The following env variables are required to make Grafana accessible via# the kubernetes api-server proxy. On production clusters, we recommend# removing these env variables, setup auth for grafana, and expose the grafana# service using a LoadBalancer or a public IP.- name: GF_AUTH_BASIC_ENABLEDvalue: "false"- name: GF_AUTH_ANONYMOUS_ENABLEDvalue: "true"- name: GF_AUTH_ANONYMOUS_ORG_ROLEvalue: Admin- name: GF_SERVER_ROOT_URL# If you're only using the API Server proxy, set this value instead:# value: /api/v1/namespaces/kube-system/services/monitoring-grafana/proxyvalue: /volumes:- name: ca-certificateshostPath:path: /etc/ssl/certs- name: grafana-storageemptyDir: {}
---
apiVersion: v1
kind: Service
metadata:labels:# For use as a Cluster add-on (https://github.com/kubernetes/kubernetes/tree/master/cluster/addons)# If you are NOT using this as an addon, you should comment out this line.kubernetes.io/cluster-service: 'true'kubernetes.io/name: monitoring-grafananame: monitoring-grafananamespace: kube-system
spec:# In a production setup, we recommend accessing Grafana through an external Loadbalancer# or through a public IP.# type: LoadBalancer# You could also use NodePort to expose the service at a randomly-generated port# type: NodePortports:- port: 80targetPort: 3000selector:k8s-app: grafanatype: NodePortkubectl apply -f grafana.yamlkubectl get pods -n kube-system -l task=monitoring -o wide
kubectl get svc -n kube-system | grep grafana  [root@master01 prometheus]# kubectl get pods -n kube-system -l task=monitoring -o wide
NAME                                 READY   STATUS    RESTARTS   AGE    IP              NODE             NOMINATED NODE   READINESS GATES
monitoring-grafana-fd8554c5c-nffvn   1/1     Running   0          3m6s   10.244.60.107   192.168.110.20   <none>           <none>
[root@master01 prometheus]# kubectl get svc -n kube-system | grep grafana           monitoring-grafana   NodePort    10.0.0.231   <none>        80:35196/TCP             3m12s

Grafana 配置

浏览器访问http://192.168.110.20:35196 ，登陆 grafana

开始配置 grafana 的 web 界面：选择 Add data source
【Name】设置成 Prometheus
【Type】选择 Prometheus
【URL】设置成 http://prometheus.monitor-sa.svc:9090 #使用service的集群内部端口配置服务端地址
点击 【Save & Test】

导入监控模板
官方链接搜索：https://grafana.com/dashboards?dataSource=prometheus&search=kubernetes

监控 node 状态
点击左侧+号选择【Import】
点击【Upload .json File】导入 node_exporter.json 模板
【Prometheus】选择 Prometheus
点击【Import】

监控 容器 状态
点击左侧+号选择【Import】
点击【Upload Upload .json File】导入 docker 模板
【Prometheus】选择 Prometheus
点击【Import】

k8s 部署 kube-state-metrics 组件
（1）安装 kube-state-metrics 组件

#创建 sa，并对 sa 授权
vim kube-state-metrics-rbac.yaml 
---
apiVersion: v1
kind: ServiceAccount
metadata:name: kube-state-metricsnamespace: kube-system
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:name: kube-state-metrics
rules:
- apiGroups: [""]resources: ["nodes", "pods", "services", "resourcequotas", "replicationcontrollers", "limitranges", "persistentvolumeclaims", "persistentvolumes", "namespaces", "endpoints"]verbs: ["list", "watch"]
- apiGroups: ["extensions"]resources: ["daemonsets", "deployments", "replicasets"]verbs: ["list", "watch"]
- apiGroups: ["apps"]resources: ["statefulsets"]verbs: ["list", "watch"]
- apiGroups: ["batch"]resources: ["cronjobs", "jobs"]verbs: ["list", "watch"]
- apiGroups: ["autoscaling"]resources: ["horizontalpodautoscalers"]verbs: ["list", "watch"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:name: kube-state-metrics
roleRef:apiGroup: rbac.authorization.k8s.iokind: ClusterRolename: kube-state-metrics
subjects:
- kind: ServiceAccountname: kube-state-metricsnamespace: kube-systemkubectl apply -f kube-state-metrics-rbac.yaml#安装 kube-state-metrics 组件和 service
vim kube-state-metrics-deploy.yaml 
---
apiVersion: apps/v1
kind: Deployment
metadata:name: kube-state-metricsnamespace: kube-system
spec:replicas: 1selector:matchLabels:app: kube-state-metricstemplate:metadata:labels:app: kube-state-metricsspec:serviceAccountName: kube-state-metricscontainers:- name: kube-state-metricsimage: quay.io/coreos/kube-state-metrics:v1.9.0ports:- containerPort: 8080
---
apiVersion: v1
kind: Service
metadata:annotations:prometheus.io/scrape: 'true'name: kube-state-metricsnamespace: kube-systemlabels:app: kube-state-metrics
spec:ports:- name: kube-state-metricsport: 8080protocol: TCPselector:app: kube-state-metricskubectl apply -f kube-state-metrics-deploy.yaml
kubectl get pods,svc -n kube-system -l app=kube-state-metrics[root@master01 prometheus]# kubectl get pods,svc -n kube-system -l app=kube-state-metrics
NAME                                      READY   STATUS    RESTARTS   AGE
pod/kube-state-metrics-58d4957bc5-2vx5j   1/1     Running   0          9sNAME                         TYPE        CLUSTER-IP   EXTERNAL-IP   PORT(S)    AGE
service/kube-state-metrics   ClusterIP   10.0.0.41    <none>        8080/TCP   9s

（2）Grafana 配置
#监控 k8s 群集状态
点击左侧+号选择【Import】
点击【Upload .json File】导入 kubernetes-cluster-prometheus_rev4.json 模板
【Prometheus】选择 Prometheus
点击【Import】

监控 k8s 群集性能状态
点击左侧+号选择【Import】
点击【Upload .json File】导入 kubernetes-cluster-monitoring-via-prometheus_rev3.json 模板
【Prometheus】选择 Prometheus
点击【Import】