简单prometheus+grafana+pushgateway采集GPU利用率和交换机流量

1、prometheus+pushgateway

从https://prometheus.io/download/下载prometheus和pushgateway

下载后修改prometheus.yaml

pushgateway需要设置honor_labels: true才能保留原来的job、instance等

然后启动：./prometheus --config.file="./prometheus.yml"

./pushgateway

访问pushgateway：http://x.x.x.x:9091

查看pushgateway数据：http://x.x.x.x:9091/metrics

注意：pushgateway只保留最后一次推送的值，prometheus拉取的时候也是拉取pushgateway当前存储的值

查看prometheus：http://x.x.x.x:9090

Status->Targets里面可以检查目标状态、上次拉取时间等

注意：http://x.x.x.x:9090/metrics并不能查看到所有表项

2、grafana

grafana安装：

apt-get install -y software-properties-common

wget -q -O - https://packages.grafana.com/gpg.key | sudo apt-key add -

add-apt-repository "deb https://packages.grafana.com/oss/deb stable main"

apt install grafana

systemctl start grafana-server

systemctl enable grafana-server

用户名密码admin admin

有的exporter源码会提供grafana配置文件，比如dcgm-exporter中：

dcgm-exporter\grafana\dcgm-exporter-dashboard.json

在Home->Dashboards->New->Import中导入json就可以看到图形

修改Y轴的单位

对应panel右上角三个点中选择Edit，Standard options，Uint中选择（忘了。。待确认）

安装dcgm-exporter

https://github.com/NVIDIA/dcgm-exporter

下载源码，然后：

kubectl create namespace gpu-monitoring

kubectl apply dcgm-exporter/service-monitor.yaml

kubectl apply dcgm-exporter/dcgm-exporter.yaml -n gpu-monitoring

创建nodeport并提交

c 复制代码

apiVersion: v1
kind: Service
metadata:
  name: dcgm-exporter
  namespace: gpu-monitoring
spec:
  selector:
    app.kubernetes.io/name: dcgm-exporter
  ports:
  - port: 9400
    targetPort: 9400
    nodePort: 30094
  type: NodePort

测试：curl http://localhost:30094/metrics

问题1：

c 复制代码

error: resource mapping not found for name: "dcgm-exporter" namespace: "" from "service-monitor.yaml": no matches for kind "ServiceMonitor" in version "monitoring.coreos.com/v1"
ensure CRDs are installed first

下载https://github.com/prometheus-operator/prometheus-operato，提交prometheus-operator\\prometheus-operator\\example\\prometheus-operator-crd\\monitoring.coreos.com_servicemonitors.yaml