253 lines
7.9 KiB
Bash
Executable File
253 lines
7.9 KiB
Bash
Executable File
#!/bin/bash
|
|
# 可观测性增强组件一键部署脚本
|
|
# 自动部署 Postgres/Redis Exporter + 更新 Prometheus/Grafana 配置
|
|
|
|
set -euo pipefail
|
|
|
|
# 颜色输出
|
|
RED='\033[0;31m'
|
|
GREEN='\033[0;32m'
|
|
YELLOW='\033[1;33m'
|
|
BLUE='\033[0;34m'
|
|
NC='\033[0m'
|
|
|
|
# 配置
|
|
MONITORING_NS="${MONITORING_NS:-monitoring}"
|
|
INFRA_NS="${INFRA_NS:-infra}"
|
|
|
|
# 辅助函数
|
|
print_header() {
|
|
echo ""
|
|
echo -e "${BLUE}========================================${NC}"
|
|
echo -e "${BLUE}$1${NC}"
|
|
echo -e "${BLUE}========================================${NC}"
|
|
echo ""
|
|
}
|
|
|
|
print_step() {
|
|
echo -e "${GREEN}==>${NC} $1"
|
|
}
|
|
|
|
print_warn() {
|
|
echo -e "${YELLOW}⚠️ $1${NC}"
|
|
}
|
|
|
|
print_error() {
|
|
echo -e "${RED}❌ $1${NC}"
|
|
}
|
|
|
|
# 检查命令
|
|
check_prerequisites() {
|
|
print_header "检查前置条件"
|
|
|
|
if ! command -v kubectl &> /dev/null; then
|
|
print_error "kubectl 未安装"
|
|
exit 1
|
|
fi
|
|
print_step "kubectl: ✓"
|
|
|
|
if ! kubectl cluster-info &> /dev/null; then
|
|
print_error "无法连接到 Kubernetes 集群"
|
|
exit 1
|
|
fi
|
|
print_step "Kubernetes 集群连接: ✓"
|
|
|
|
# 检查核心组件
|
|
if ! kubectl get deployment prometheus -n "$MONITORING_NS" &> /dev/null; then
|
|
print_error "Prometheus 未部署在 namespace $MONITORING_NS"
|
|
exit 1
|
|
fi
|
|
print_step "Prometheus: ✓"
|
|
|
|
if ! kubectl get deployment grafana -n "$MONITORING_NS" &> /dev/null; then
|
|
print_warn "Grafana 未部署，将跳过 Grafana 配置"
|
|
else
|
|
print_step "Grafana: ✓"
|
|
fi
|
|
}
|
|
|
|
# 部署 Exporters
|
|
deploy_exporters() {
|
|
print_header "部署 Exporters"
|
|
|
|
# Postgres Exporter
|
|
read -p "是否部署 Postgres Exporter? (y/N): " deploy_pg
|
|
if [[ $deploy_pg =~ ^[Yy]$ ]]; then
|
|
print_step "部署 Postgres Exporter..."
|
|
|
|
# 检查 Secret 是否存在
|
|
if kubectl get secret postgres-exporter-secret -n "$MONITORING_NS" &> /dev/null; then
|
|
print_warn "postgres-exporter-secret 已存在，跳过创建"
|
|
else
|
|
print_warn "请先编辑 infra/observability/postgres-secret-example.yaml 配置连接串"
|
|
read -p "按回车继续，或 Ctrl+C 取消..."
|
|
|
|
kubectl apply -f infra/observability/postgres-secret-example.yaml
|
|
fi
|
|
|
|
kubectl apply -f infra/observability/postgres-exporter.yaml
|
|
print_step "Postgres Exporter 部署完成"
|
|
fi
|
|
|
|
# Redis Exporter
|
|
read -p "是否部署 Redis Exporter? (y/N): " deploy_redis
|
|
if [[ $deploy_redis =~ ^[Yy]$ ]]; then
|
|
print_step "部署 Redis Exporter..."
|
|
|
|
if kubectl get secret redis-exporter-secret -n "$MONITORING_NS" &> /dev/null; then
|
|
print_warn "redis-exporter-secret 已存在，跳过创建"
|
|
else
|
|
print_warn "请先编辑 infra/observability/redis-secret-example.yaml 配置连接串"
|
|
read -p "按回车继续，或 Ctrl+C 取消..."
|
|
|
|
kubectl apply -f infra/observability/redis-secret-example.yaml
|
|
fi
|
|
|
|
kubectl apply -f infra/observability/redis-exporter.yaml
|
|
print_step "Redis Exporter 部署完成"
|
|
fi
|
|
}
|
|
|
|
# 更新 Prometheus 配置
|
|
update_prometheus() {
|
|
print_header "更新 Prometheus 配置"
|
|
|
|
read -p "是否更新 Prometheus ConfigMap? (Y/n): " update_prom
|
|
if [[ ! $update_prom =~ ^[Nn]$ ]]; then
|
|
print_step "备份现有 Prometheus ConfigMap..."
|
|
kubectl get configmap prometheus-config -n "$MONITORING_NS" -o yaml > \
|
|
"prometheus-config-backup-$(date +%Y%m%d-%H%M%S).yaml"
|
|
|
|
print_step "应用新的 Prometheus 配置..."
|
|
kubectl apply -f monitoring/prometheus/prometheus-configmap-enhanced.yaml
|
|
|
|
print_step "部署告警规则..."
|
|
|
|
# 创建规则 ConfigMaps
|
|
for rule_file in monitoring/prometheus/rules/*.yml; do
|
|
rule_name=$(basename "$rule_file" .yml)
|
|
print_step "部署规则: $rule_name"
|
|
|
|
kubectl create configmap "prometheus-rules-$rule_name" \
|
|
--from-file="$rule_file" \
|
|
-n "$MONITORING_NS" \
|
|
--dry-run=client -o yaml | kubectl apply -f -
|
|
done
|
|
|
|
print_step "重启 Prometheus 以加载新配置..."
|
|
kubectl rollout restart deployment/prometheus -n "$MONITORING_NS"
|
|
|
|
print_step "等待 Prometheus 就绪..."
|
|
kubectl rollout status deployment/prometheus -n "$MONITORING_NS" --timeout=120s
|
|
fi
|
|
}
|
|
|
|
# 更新 Grafana 配置
|
|
update_grafana() {
|
|
print_header "更新 Grafana 配置"
|
|
|
|
if ! kubectl get deployment grafana -n "$MONITORING_NS" &> /dev/null; then
|
|
print_warn "Grafana 未部署，跳过"
|
|
return
|
|
fi
|
|
|
|
read -p "是否更新 Grafana 数据源配置? (Y/n): " update_grafana
|
|
if [[ ! $update_grafana =~ ^[Nn]$ ]]; then
|
|
print_step "备份现有 Grafana 数据源配置..."
|
|
kubectl get configmap grafana-datasources -n "$MONITORING_NS" -o yaml > \
|
|
"grafana-datasources-backup-$(date +%Y%m%d-%H%M%S).yaml"
|
|
|
|
print_step "应用新的数据源配置..."
|
|
kubectl apply -f monitoring/grafana/datasources-loki-tempo-enhanced.yaml
|
|
|
|
print_step "重启 Grafana 以加载新配置..."
|
|
kubectl rollout restart deployment/grafana -n "$MONITORING_NS"
|
|
|
|
print_step "等待 Grafana 就绪..."
|
|
kubectl rollout status deployment/grafana -n "$MONITORING_NS" --timeout=120s
|
|
fi
|
|
}
|
|
|
|
# 导入 Dashboards
|
|
import_dashboards() {
|
|
print_header "导入 Grafana Dashboards"
|
|
|
|
if ! kubectl get deployment grafana -n "$MONITORING_NS" &> /dev/null; then
|
|
print_warn "Grafana 未部署，跳过"
|
|
return
|
|
fi
|
|
|
|
read -p "是否导入 Grafana Dashboards? (y/N): " import_dash
|
|
if [[ $import_dash =~ ^[Yy]$ ]]; then
|
|
print_step "执行 Dashboard 导入脚本..."
|
|
cd monitoring/grafana/dashboards
|
|
./import-dashboards.sh
|
|
cd -
|
|
fi
|
|
}
|
|
|
|
# 验收
|
|
run_verification() {
|
|
print_header "运行验收测试"
|
|
|
|
read -p "是否运行验收测试? (Y/n): " run_verify
|
|
if [[ ! $run_verify =~ ^[Nn]$ ]]; then
|
|
print_step "执行验收脚本..."
|
|
./verify-observability-enhanced.sh
|
|
fi
|
|
}
|
|
|
|
# 显示后续步骤
|
|
show_next_steps() {
|
|
print_header "部署完成！"
|
|
|
|
echo -e "${GREEN}✅ 可观测性增强组件部署成功${NC}"
|
|
echo ""
|
|
echo "后续步骤："
|
|
echo ""
|
|
echo "1. 验证 Prometheus Targets:"
|
|
echo " kubectl port-forward -n $MONITORING_NS svc/prometheus 9090:9090"
|
|
echo " 浏览器打开: http://localhost:9090/targets"
|
|
echo ""
|
|
echo "2. 访问 Grafana Dashboards:"
|
|
echo " kubectl port-forward -n $MONITORING_NS svc/grafana 3000:3000"
|
|
echo " 浏览器打开: http://localhost:3000"
|
|
echo ""
|
|
echo "3. 启用 pg_stat_statements (如果使用 Postgres):"
|
|
echo " 参考文档: docs/observability/pg-stat-statements-setup.md"
|
|
echo ""
|
|
echo "4. 集成业务应用:"
|
|
echo " - Spring Boot: docs/observability/business-instrumentation-spring-boot.md"
|
|
echo " - Python: docs/observability/business-instrumentation-python.md"
|
|
echo ""
|
|
echo "5. 测试 Loki ↔ Tempo 跳转:"
|
|
echo " - 在 Grafana Explore 中查询包含 trace_id 的日志"
|
|
echo " - 点击 'View Trace' 链接验证跳转"
|
|
echo ""
|
|
echo "📚 完整文档: OBSERVABILITY-ENHANCEMENT-GUIDE.md"
|
|
echo ""
|
|
}
|
|
|
|
# 主函数
|
|
main() {
|
|
echo -e "${BLUE}"
|
|
cat << "EOF"
|
|
╔═══════════════════════════════════════════════════════╗
|
|
║ K8s 可观测性增强部署工具 ║
|
|
║ Postgres/Redis Monitoring + LGTM Integration ║
|
|
╚═══════════════════════════════════════════════════════╝
|
|
EOF
|
|
echo -e "${NC}"
|
|
|
|
check_prerequisites
|
|
deploy_exporters
|
|
update_prometheus
|
|
update_grafana
|
|
import_dashboards
|
|
run_verification
|
|
show_next_steps
|
|
}
|
|
|
|
# 执行
|
|
main "$@"
|