KnowledgeBase/docker/check-services.sh

#!/bin/bash

# Docker服务健康检查脚本
# 用于检查所有服务的健康状态

set -e

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

# 日志函数
log_info() {
    echo -e "${BLUE}[INFO]${NC} $1"
}

log_success() {
    echo -e "${GREEN}[SUCCESS]${NC} $1"
}

log_warning() {
    echo -e "${YELLOW}[WARNING]${NC} $1"
}

log_error() {
    echo -e "${RED}[ERROR]${NC} $1"
}

# 检查Docker是否运行
check_docker() {
    log_info "检查Docker服务状态..."
    if ! docker info >/dev/null 2>&1; then
        log_error "Docker未运行或无法访问"
        exit 1
    fi
    log_success "Docker服务正常运行"
}

# 检查容器状态
check_container_status() {
    local container_name=$1
    local service_name=$2
    
    log_info "检查${service_name}容器状态..."
    
    if ! docker ps --format "table {{.Names}}" | grep -q "^${container_name}$"; then
        log_error "${service_name}容器未运行"
        return 1
    fi
    
    # 检查容器健康状态
    local health_status=$(docker inspect --format='{{.State.Health.Status}}' ${container_name} 2>/dev/null || echo "no-healthcheck")
    
    case $health_status in
        "healthy")
            log_success "${service_name}容器健康状态正常"
            return 0
            ;;
        "unhealthy")
            log_error "${service_name}容器健康检查失败"
            return 1
            ;;
        "starting")
            log_warning "${service_name}容器正在启动中..."
            return 2
            ;;
        "no-healthcheck")
            log_warning "${service_name}容器未配置健康检查"
            return 0
            ;;
        *)
            log_warning "${service_name}容器健康状态未知: ${health_status}"
            return 0
            ;;
    esac
}

# 检查MySQL数据库连接
check_mysql() {
    log_info "检查MySQL数据库连接..."
    
    local max_attempts=5
    local attempt=1
    
    while [ $attempt -le $max_attempts ]; do
        if docker exec knowledge_base_mysql mysqladmin ping -h localhost --silent 2>/dev/null; then
            log_success "MySQL数据库连接正常"
            return 0
        fi
        
        log_warning "MySQL连接尝试 ${attempt}/${max_attempts} 失败，等待5秒后重试..."
        sleep 5
        ((attempt++))
    done
    
    log_error "MySQL数据库连接失败"
    return 1
}

# 检查Redis连接
check_redis() {
    log_info "检查Redis缓存连接..."
    
    if docker exec knowledge_base_redis redis-cli ping | grep -q "PONG"; then
        log_success "Redis缓存连接正常"
        return 0
    else
        log_error "Redis缓存连接失败"
        return 1
    fi
}

# 检查Meilisearch连接
check_meilisearch() {
    log_info "检查Meilisearch搜索引擎连接..."
    
    local max_attempts=3
    local attempt=1
    
    while [ $attempt -le $max_attempts ]; do
        if docker exec knowledge_base_meilisearch curl -f http://localhost:7700/health >/dev/null 2>&1; then
            log_success "Meilisearch搜索引擎连接正常"
            return 0
        fi
        
        log_warning "Meilisearch连接尝试 ${attempt}/${max_attempts} 失败，等待3秒后重试..."
        sleep 3
        ((attempt++))
    done
    
    log_error "Meilisearch搜索引擎连接失败"
    return 1
}

# 检查Web应用健康状态
check_web_app() {
    log_info "检查Web应用健康状态..."
    
    local max_attempts=3
    local attempt=1
    
    while [ $attempt -le $max_attempts ]; do
        local response=$(curl -s -o /dev/null -w "%{http_code}" http://localhost:8000/health 2>/dev/null || echo "000")
        
        if [ "$response" = "200" ]; then
            log_success "Web应用健康检查通过"
            return 0
        elif [ "$response" = "503" ]; then
            log_warning "Web应用部分服务不可用，但应用仍在运行"
            return 2
        fi
        
        # 如果没有专门的健康检查路由，尝试访问根路径
        local root_response=$(curl -s -o /dev/null -w "%{http_code}" http://localhost:8000/ 2>/dev/null || echo "000")
        if [ "$root_response" = "200" ]; then
            log_success "Web应用根路径访问正常"
            return 0
        fi
        
        log_warning "Web应用健康检查尝试 ${attempt}/${max_attempts} 失败 (HTTP: ${response})，等待5秒后重试..."
        sleep 5
        ((attempt++))
    done
    
    log_error "Web应用健康检查失败"
    return 1
}

# 检查队列处理器
check_queue_worker() {
    log_info "检查队列处理器状态..."
    
    # 检查应用容器中的队列进程是否正在运行
    if docker exec knowledge_base_app pgrep -f "queue:work" >/dev/null 2>&1; then
        log_success "队列处理器正常运行"
        return 0
    else
        log_error "队列处理器进程未运行"
        return 1
    fi
}

# 检查数据持久化
check_data_persistence() {
    log_info "检查数据持久化状态..."
    
    local errors=0
    
    # 检查存储目录
    local storage_dirs=("./storage/mysql" "./storage/redis" "./storage/meilisearch" "./storage/app")
    
    for dir in "${storage_dirs[@]}"; do
        if [ ! -d "$dir" ]; then
            log_error "存储目录不存在: $dir"
            ((errors++))
        elif [ ! -w "$dir" ]; then
            log_error "存储目录不可写: $dir"
            ((errors++))
        fi
    done
    
    if [ $errors -eq 0 ]; then
        log_success "数据持久化配置正常"
        return 0
    else
        log_error "发现 $errors 个数据持久化问题"
        return 1
    fi
}

# 主检查函数
main() {
    echo "========================================"
    echo "Docker服务健康检查开始"
    echo "时间: $(date)"
    echo "========================================"
    
    local total_checks=0
    local failed_checks=0
    local warning_checks=0
    
    # 执行所有检查
    checks=(
        "check_docker:Docker服务"
        "check_container_status:knowledge_base_mysql:MySQL容器"
        "check_container_status:knowledge_base_redis:Redis容器"
        "check_container_status:knowledge_base_meilisearch:Meilisearch容器"
        "check_container_status:knowledge_base_app:应用容器"
        "check_mysql:MySQL连接"
        "check_redis:Redis连接"
        "check_meilisearch:Meilisearch连接"
        "check_web_app:Web应用"
        "check_queue_worker:队列处理器"
        "check_data_persistence:数据持久化"
    )
    
    for check in "${checks[@]}"; do
        IFS=':' read -ra CHECK_PARTS <<< "$check"
        local check_func="${CHECK_PARTS[0]}"
        local check_args=("${CHECK_PARTS[@]:1}")
        
        ((total_checks++))
        
        if [ ${#check_args[@]} -eq 0 ]; then
            $check_func
        else
            $check_func "${check_args[@]}"
        fi
        
        local result=$?
        if [ $result -eq 1 ]; then
            ((failed_checks++))
        elif [ $result -eq 2 ]; then
            ((warning_checks++))
        fi
        
        echo ""
    done
    
    # 输出总结
    echo "========================================"
    echo "健康检查完成"
    echo "总检查项: $total_checks"
    echo "失败: $failed_checks"
    echo "警告: $warning_checks"
    echo "成功: $((total_checks - failed_checks - warning_checks))"
    echo "========================================"
    
    if [ $failed_checks -gt 0 ]; then
        log_error "发现 $failed_checks 个严重问题，请检查服务状态"
        exit 1
    elif [ $warning_checks -gt 0 ]; then
        log_warning "发现 $warning_checks 个警告，服务可能需要关注"
        exit 2
    else
        log_success "所有服务健康检查通过"
        exit 0
    fi
}

# 如果脚本被直接执行
if [[ "${BASH_SOURCE[0]}" == "${0}" ]]; then
    main "$@"
fi