infra-deploy-guide

name	infra-deploy-guide
description	Infrastructure deployment guide. For Kubernetes cluster initialization, ArgoCD deployment, VictoriaMetrics monitoring stack deployment, VictoriaLogs logging stack deployment, kube-prometheus-stack metrics collection deployment. 触发场景：集群初始化、基础设施部署、监控系统搭建、日志系统搭建、ArgoCD GitOps 配置。
allowed-tools	helm_deploy helm_status kubectl_get kubectl_apply create_namespace update_helm_values create_argocd_app_manifest git_commit_and_push

Infrastructure Deployment Guide

Cluster Initialization Deployment Flow

Prerequisite Check

1. Confirm target cluster kubeconfig is reachable
2. Confirm Helm repositories are configured
3. Confirm GitOps repository is accessible

Deployment Order (Dependencies)

1. ArgoCD          → Base, other components deployed via ArgoCD GitOps
2. VictoriaMetrics → Metrics storage backend
3. kube-prometheus-stack → Metrics collection (remote_write to VM)
4. VictoriaLogs    → Log storage backend

Step 1: Deploy ArgoCD

# Create namespace
kubectl create namespace argocd

# Deploy with Helm
helm install argocd argo/argo-cd \
  --namespace argocd \
  -f charts/argocd/values.yaml

Key config items (charts/argocd/values.yaml):

• server.ingress.enabled: Enable Ingress
• server.ingress.hosts: Access domain
• configs.repositories: Git repository config

Verification:

• ArgoCD Server Pod Running
• ArgoCD Repo Server Pod Running
• UI/CLI login works

Step 2: Deploy VictoriaMetrics

helm install victoria-metrics vm/victoria-metrics-k8s-stack \
  --namespace monitoring --create-namespace \
  -f charts/victoria-metrics-stack/values.yaml

Key config items:

• vmsingle.spec.retentionPeriod: Data retention (default 30d)
• vmsingle.spec.storage.storageClassName: Storage class
• vmsingle.spec.storage.resources.requests.storage: Storage size

Verification:

• VMSingle/VMCluster Pod Running
• VMAlert Pod Running (if enabled)
• API can query metrics

Step 3: Deploy kube-prometheus-stack

helm install kps prometheus-community/kube-prometheus-stack \
  --namespace monitoring \
  -f charts/kube-prometheus-stack/values.yaml

Key config:

• prometheus.prometheusSpec.remoteWrite: Configure remote_write to VictoriaMetrics
• Disable Grafana (use standalone or VM built-in)
• Disable AlertManager (use VMAlert)

Verification:

• Prometheus Pod Running
• node-exporter DaemonSet Running
• kube-state-metrics Running
• remote_write target reachable

Step 4: Deploy VictoriaLogs

helm install victoria-logs vm/victoria-logs-single \
  --namespace logging --create-namespace \
  -f charts/victoria-logs/values.yaml

Key config:

• server.retentionPeriod: Log retention
• server.persistentVolume.size: Storage size
• Log collector (fluent-bit/vector) config

Verification:

• VictoriaLogs Pod Running
• Log collector DaemonSet Running
• LogsQL can query logs

Post-Deployment Verification Checklist

Component	Check Item	Method
ArgoCD	Server running	kubectl get pods -n argocd
ArgoCD	UI accessible	Browser to Ingress URL
VictoriaMetrics	VMSingle running	kubectl get pods -n monitoring -l app=vmsingle
VictoriaMetrics	Metrics writing	PromQL: up
KPS	Prometheus running	kubectl get pods -n monitoring -l app.kubernetes.io/name=prometheus
KPS	remote_write OK	Prometheus targets page
VictoriaLogs	Server running	kubectl get pods -n logging
VictoriaLogs	Log writing	LogsQL: *

Troubleshooting

ArgoCD Deployment Failure

• Check PVC binding (requires StorageClass)
• Check Ingress Controller exists
• Check Git repo connectivity (SSH key / token)

VictoriaMetrics Deployment Failure

• Check storage quota is sufficient
• Check CRD installed (vm-operator)

remote_write Not Working

• Check VictoriaMetrics endpoint address
• Check network policies allow cross-namespace traffic
• Check auth configuration