1. 部署

1. 部署

1.1 下载源码

git clone https://github.com/prometheus-operator/kube-prometheus.git
cd kube-prometheus

注意：请确认好**Kubernetes**集群版本，下载对应版本的源码包。

1.2 修改yaml

主要是修改一下Service的类型和镜像源。默认Service为ClusterIP，为了提供一个集群访问入口，这里设置为NodeProt类型，如果是在云服务上部署的，可以改成LoadBalancer；另外就是有几个服务的镜像源为谷歌的镜像仓库，国内无法访问，因此需要更换一下镜像。

1.2.1 修改`grafana-service.yaml`

sed -i '/spec:/a\  type: NodePort' manifests/grafana-service.yaml

1.2.2 修改`prometheus-service.yaml`

sed -i '/spec:/a\  type: NodePort' manifests/prometheus-service.yaml

1.2.3 修改`alertmanager-service.yaml`

sed -i '/spec:/a\  type: NodePort' manifests/alertmanager-service.yaml

1.2.4 修改镜像源

查看包含image的yaml文件：grep -l image: manifests/*.yaml

sed -ri 's#(image:) k8s.*#\1 bitnami/kube-state-metrics:2.4.2#' manifests/kubeStateMetrics-deployment.yaml
sed -ri 's#(image:) k8s.*#\1 willdockerhub/prometheus-adapter:v0.9.1#' manifests/prometheusAdapter-deployment.yaml

1.3 部署 CRD

kubectl create -f manifests/setup

注意：使用**apply**时会提示以下错误，可以使用**create**代替。

The CustomResourceDefinition "prometheuses.monitoring.coreos.com" is invalid: metadata.annotations: Too long: must have at most 262144 bytes

1.4 部署 Prometheus 生态组件

注意：默认为高可用部署，即关键组件都会启动多个副本，副本数请按需修改。

kubectl apply -f manifests

查看部署情况：

[root@k8s-master01 ~]# kubectl get pod -n monitoring
NAME                                   READY   STATUS    RESTARTS   AGE
alertmanager-main-0                    2/2     Running   0          35m
blackbox-exporter-5cb5d7479d-n59tf     3/3     Running   0          42m
grafana-d595885ff-6gg6t                1/1     Running   0          42m
kube-state-metrics-d64589d79-hwx7j     3/3     Running   0          42m
node-exporter-9l7k5                    2/2     Running   0          42m
node-exporter-fm78w                    2/2     Running   0          42m
node-exporter-tw54d                    2/2     Running   0          42m
prometheus-adapter-86df476d87-27dp4    1/1     Running   0          42m
prometheus-adapter-86df476d87-65dfz    1/1     Running   0          42m
prometheus-k8s-0                       2/2     Running   0          35m
prometheus-operator-7cc5c6fd68-xfkfh   2/2     Running   0          42m
[root@k8s-master01 ~]# kubectl get svc -n monitoring
NAME                    TYPE        CLUSTER-IP       EXTERNAL-IP   PORT(S)                         AGE
alertmanager-main       ClusterIP   10.109.19.210    <none>        9093/TCP,8080/TCP               64m
alertmanager-operated   ClusterIP   None             <none>        9093/TCP,9094/TCP,9094/UDP      57m
blackbox-exporter       ClusterIP   10.99.75.89      <none>        9115/TCP,19115/TCP              64m
grafana                 NodePort    10.99.120.219    <none>        3000:32287/TCP                  64m
kube-state-metrics      ClusterIP   None             <none>        8443/TCP,9443/TCP               64m
node-exporter           ClusterIP   None             <none>        9100/TCP                        64m
prometheus-adapter      ClusterIP   10.103.205.208   <none>        443/TCP                         64m
prometheus-k8s          NodePort    10.99.26.234     <none>        9090:30447/TCP,8080:30361/TCP   64m
prometheus-operated     ClusterIP   None             <none>        9090/TCP                        57m
prometheus-operator     ClusterIP   None             <none>        8443/TCP                        64m

kube-prometheus