#!/bin/bash
#
# 节点健康检查脚本
# 使用方法: bash check-node-health.sh
#

# 颜色输出
RED='\033[0;31m'
GREEN='\033[0;32m'
YELLOW='\033[1;33m'
BLUE='\033[0;34m'
NC='\033[0m' # No Color

echo -e "${BLUE}================================${NC}"
echo -e "${BLUE}K3s 集群健康检查${NC}"
echo -e "${BLUE}================================${NC}"
echo ""

# 1. 检查节点状态
echo -e "${YELLOW}[1/8] 检查节点状态...${NC}"
kubectl get nodes -o wide
echo ""

# 2. 检查节点资源
echo -e "${YELLOW}[2/8] 检查节点资源使用...${NC}"
kubectl top nodes 2>/dev/null || echo -e "${YELLOW}⚠ metrics-server 未就绪${NC}"
echo ""

# 3. 检查系统 Pods
echo -e "${YELLOW}[3/8] 检查系统组件...${NC}"
kubectl get pods -n kube-system
echo ""

# 4. 检查 Longhorn
echo -e "${YELLOW}[4/8] 检查 Longhorn 存储...${NC}"
kubectl get pods -n longhorn-system | head -10
echo ""

# 5. 检查 PVC
echo -e "${YELLOW}[5/8] 检查持久化存储卷...${NC}"
kubectl get pvc -A
echo ""

# 6. 检查应用 Pods
echo -e "${YELLOW}[6/8] 检查应用 Pods...${NC}"
kubectl get pods -A | grep -v "kube-system\|longhorn-system\|cert-manager" | head -20
echo ""

# 7. 检查 Ingress
echo -e "${YELLOW}[7/8] 检查 Ingress 配置...${NC}"
kubectl get ingress -A
echo ""

# 8. 检查证书
echo -e "${YELLOW}[8/8] 检查 SSL 证书...${NC}"
kubectl get certificate -A
echo ""

# 统计信息
echo -e "${BLUE}================================${NC}"
echo -e "${BLUE}集群统计信息${NC}"
echo -e "${BLUE}================================${NC}"

TOTAL_NODES=$(kubectl get nodes --no-headers | wc -l)
READY_NODES=$(kubectl get nodes --no-headers | grep " Ready " | wc -l)
TOTAL_PODS=$(kubectl get pods -A --no-headers | wc -l)
RUNNING_PODS=$(kubectl get pods -A --no-headers | grep "Running" | wc -l)
TOTAL_PVC=$(kubectl get pvc -A --no-headers | wc -l)
BOUND_PVC=$(kubectl get pvc -A --no-headers | grep "Bound" | wc -l)

echo -e "节点总数: ${GREEN}${TOTAL_NODES}${NC} (就绪: ${GREEN}${READY_NODES}${NC})"
echo -e "Pod 总数: ${GREEN}${TOTAL_PODS}${NC} (运行中: ${GREEN}${RUNNING_PODS}${NC})"
echo -e "PVC 总数: ${GREEN}${TOTAL_PVC}${NC} (已绑定: ${GREEN}${BOUND_PVC}${NC})"
echo ""

# 健康评分
if [ $READY_NODES -eq $TOTAL_NODES ] && [ $RUNNING_PODS -gt $((TOTAL_PODS * 80 / 100)) ]; then
    echo -e "${GREEN}✓ 集群健康状态: 良好${NC}"
elif [ $READY_NODES -gt $((TOTAL_NODES / 2)) ]; then
    echo -e "${YELLOW}⚠ 集群健康状态: 一般${NC}"
else
    echo -e "${RED}✗ 集群健康状态: 异常${NC}"
fi
echo ""