Middleware
监控与观测
自动化运维
CI/CD&GitOps环境前提
- 已配置
kubeconfig,可以执行kubectl - 目标节点至少有 50 GiB 可用磁盘空间
- 目标节点可以拉取 VictoriaLogs 镜像
本清单使用节点本地
hostPath 持久卷,并将 Pod 固定在 worko1 节点。应用清单前,请把节点名替换为实际值。生产环境请使用 NFS、Ceph、Longhorn 或云厂商 CSI 存储。kubectl get nodes
部署参数
| 配置 | 默认值 | 说明 |
|---|---|---|
| 镜像 | uhub.service.ucloud.cn/kubernetes_images/victoria-logs:latest | 生产环境应固定版本 |
| 数据目录 | worko1:/var/lib/observability/victorialogs | 节点本地目录 |
| 存储容量 | 50Gi | 按日志量和保留周期调整 |
| 保留周期 | 30d | 由 -retentionPeriod 控制 |
| NodePort | 30942 | 对外提供 HTTP 接口 |
部署清单
把以下内容保存为victorialogs.yaml:
victorialogs.yaml
apiVersion: v1
kind: Namespace
metadata:
name: observability
labels:
app.kubernetes.io/part-of: observability
---
apiVersion: v1
kind: PersistentVolume
metadata:
name: victorialogs-local-pv
labels:
app.kubernetes.io/name: victorialogs
spec:
capacity:
storage: 50Gi
accessModes:
- ReadWriteOnce
persistentVolumeReclaimPolicy: Retain
storageClassName: ""
hostPath:
path: /var/lib/observability/victorialogs
type: DirectoryOrCreate
nodeAffinity:
required:
nodeSelectorTerms:
- matchExpressions:
- key: kubernetes.io/hostname
operator: In
values:
- worko1
---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
name: victorialogs-data
namespace: observability
spec:
accessModes:
- ReadWriteOnce
storageClassName: ""
volumeName: victorialogs-local-pv
resources:
requests:
storage: 50Gi
---
apiVersion: v1
kind: Service
metadata:
name: victorialogs
namespace: observability
labels:
app.kubernetes.io/name: victorialogs
spec:
type: NodePort
selector:
app.kubernetes.io/name: victorialogs
ports:
- name: http
port: 9428
targetPort: http
nodePort: 30942
---
apiVersion: apps/v1
kind: StatefulSet
metadata:
name: victorialogs
namespace: observability
labels:
app.kubernetes.io/name: victorialogs
spec:
serviceName: victorialogs
replicas: 1
selector:
matchLabels:
app.kubernetes.io/name: victorialogs
template:
metadata:
labels:
app.kubernetes.io/name: victorialogs
spec:
nodeSelector:
kubernetes.io/hostname: worko1
containers:
- name: victorialogs
image: uhub.service.ucloud.cn/kubernetes_images/victoria-logs:latest
imagePullPolicy: IfNotPresent
args:
- -storageDataPath=/victoria-logs-data
- -retentionPeriod=30d
- -httpListenAddr=:9428
ports:
- name: http
containerPort: 9428
readinessProbe:
httpGet:
path: /health
port: http
initialDelaySeconds: 10
periodSeconds: 10
livenessProbe:
httpGet:
path: /health
port: http
initialDelaySeconds: 30
periodSeconds: 20
resources:
requests:
cpu: 250m
memory: 512Mi
limits:
cpu: "2"
memory: 2Gi
volumeMounts:
- name: data
mountPath: /victoria-logs-data
volumes:
- name: data
persistentVolumeClaim:
claimName: victorialogs-data
应用并验证
kubectl apply -f victorialogs.yaml
kubectl -n observability rollout status statefulset/victorialogs --timeout=180s
kubectl -n observability get pod,pvc,service -l app.kubernetes.io/name=victorialogs
<节点IP> 替换为集群节点地址,检查服务健康状态:
curl http://<节点IP>:30942/health
OK。
写入与查询测试
写入一条测试日志:curl -X POST http://<节点IP>:30942/insert/jsonline \
-H 'Content-Type: application/stream+json' \
--data-binary '{"_time":"2026-09-10T03:00:00Z","_msg":"observability test","level":"info","service":"demo"}'
curl --get 'http://<节点IP>:30942/select/logsql/query' --data-urlencode 'query=_time:1h'
curl --get 'http://<节点IP>:30942/select/logsql/query' --data-urlencode 'query=service:demo'
常见故障
| 现象 | 排查方法 |
|---|---|
| Pod 一直 Pending | 检查 worko1 节点名、PV 的 nodeAffinity 和 PVC 绑定状态 |
| Pod 镜像拉取失败 | 执行 kubectl -n observability describe pod victorialogs-0,检查镜像仓库连通性 |
| 健康检查失败 | 查看 kubectl -n observability logs victorialogs-0,确认数据目录可写 |
| 查询不到日志 | 先手工写入测试日志,再检查 Vector 的 sink 状态 |
删除组件
kubectl -n observability delete statefulset victorialogs
kubectl -n observability delete service victorialogs
kubectl -n observability delete pvc victorialogs-data
PV 的回收策略是
Retain。删除 PVC 不会删除 PV 和节点数据。确认数据不再需要后,再删除 victorialogs-local-pv 并清理节点目录。下一步
- 部署 Vector:采集 Kubernetes 容器日志
- 部署 Grafana:查询和展示日志