1.部署prometheus

创建存储类并设置默认

helm repo add nfs-subdir-external-provisioner https://kubernetes-sigs.github.io/nfs-subdir-external-provisioner/
helm repo update

#前置需要部署NFS 搜索站内nfs教程
helm install nfs-provisioner nfs-subdir-external-provisioner/nfs-subdir-external-provisioner \
  --set nfs.server=192.168.163.139 \
  --set nfs.path=/data/share

kubectl patch storageclass nfs-client \
  -p '{"metadata": {"annotations":{"storageclass.kubernetes.io/is-default-class":"true"}}}'

添加 prometheus-community 仓库

helm repo add prometheus-community https://prometheus-community.github.io/helm-charts
helm repo update

创建监控命名空间

kubectl create namespace monitoring

部署 prometheus（不带 Grafana 方便Grafana独立后续接入其他数据源）

helm install prometheus prometheus-community/prometheus \
  --namespace monitoring

接下来我们就可以继续部署 Loki（日志收集） 和 Grafana（可视化）。
我会给你一套完整、可直接执行的命令。

---

2. 安装 Loki-stack（grafana+loki+promtail 日志系统）

1. 添加 Helm 仓库

helm repo add grafana https://grafana.github.io/helm-charts
helm repo update

编辑安装配置

vim loki-stack-values.yaml 

grafana:
  enabled: true
  service:
    type: NodePort
    nodePort: 31200 #grafana 访问端口,自行设置

loki:
  persistence:
    enabled: true
    storageClassName: nfs-client
    size: 5Gi #自行设置大小,不要超过nfs大小
  tableManager:
    retention_deletes_enabled: true
    retention_period: 168h  # 日志保存时长自行设置

promtail:
  enabled: true
  serviceAccount:
    create: true
    name: loki-promtail
    clusterRole:
      enabled: true
      rules:
        - apiGroups: [""]
          resources: ["pods", "namespaces"]
          verbs: ["get","list","watch"]
        - apiGroups: [""]
          resources: ["pods/log"]
          verbs: ["get","list","watch"]

  config:
    clients:
      - url: http://loki:3100/loki/api/v1/push

    positions:
      filename: /run/promtail/positions.yaml

    scrape_configs:
      - job_name: kubernetes-pods
        kubernetes_sd_configs:
          - role: pod
            namespaces: {}   # 空对象表示所有命名空间
        relabel_configs:
          - source_labels: [__meta_kubernetes_namespace]
            target_label: namespace
          - source_labels: [__meta_kubernetes_pod_name]
            target_label: pod
          - source_labels: [__meta_kubernetes_pod_container_name]
            target_label: container
          - source_labels: [__meta_kubernetes_pod_label_app]
            target_label: app

安装

helm upgrade --install loki grafana/loki-stack \
  -n monitoring --create-namespace \
  -f loki-stack-values.yaml

获取grafana密码

登陆默认有loki数据源
prometheus 额外添加
URL: http://prometheus-server.monitoring.svc.cluster.local

 kubectl get secret --namespace monitoring loki-grafana -o jsonpath="{.data.admin-password}" | base64 --decode

常见问题:

有个镜像拉取超时失败,手动拉取

ctr -n k8s.io images pull docker.io/kiwigrid/k8s-sidecar:1.19.2

孤儿pod产生

由于kuboard面板操作过东西
卸载后

root@u1:~/k8s_yaml# helm uninstall loki -n monitoring 
release "loki" uninstalled
root@u1:~/k8s_yaml# kubectl get pod -n monitoring 
NAME                                                READY   STATUS    RESTARTS   AGE
loki-grafana-bcd96655d-hbps5                        2/2     Running   0          18h 

检查

• Grafana 部署仍在；

• 但 Helm 已丢失这个 release 的记录（即 “孤儿资源”）；

• 所以 helm uninstall loki 已经无法再回收这些对象。

  root@u1:~/k8s_yaml# helm list -n monitoring
  NAME            NAMESPACE       REVISION        UPDATED                                 STATUS          CHART                   APP VERSION
  prometheus      monitoring      1               2025-09-24 07:49:58.007039569 +0000 UTC deployed        prometheus-27.39.0      v3.6.0     
  root@u1:~/k8s_yaml# kubectl get deployment loki-grafana -n monitoring -o jsonpath='{.metadata.labels}'
  {"app.kubernetes.io/instance":"loki","app.kubernetes.io/managed-by":"Helm","app.kubernetes.io/name":"grafana","app.kubernetes.io/version":"10.3.3","helm.sh/chart":"grafana-6.43.5","k8s.kuboard.cn/name":"loki-grafana"}root@u1:~/k8s_yaml# 

  孤儿pod删除

  kubectl delete deployment loki-grafana -n monitoring
  kubectl delete svc loki-grafana -n monitoring
  kubectl delete secret loki-grafana -n monitoring --ignore-not-found
  kubectl delete configmap loki-grafana -n monitoring --ignore-not-found
  kubectl delete pvc -l app.kubernetes.io/instance=loki -n monitoring --ignore-not-found