backend

幂等性设计模式：构建可靠的分布式系统

深入讲解幂等性在分布式系统中的重要性，涵盖幂等键、状态机、Token机制等核心模式，详解数据库唯一约束、请求去重、重试策略等实战技巧，提供完整的Go语言实现代码。

spcent 2026-09-08 4 分钟阅读 1950 字

引言

在网络不可靠的分布式环境中，请求可能因为超时、重试、消息重复投递等原因被执行多次。幂等性设计确保同一操作执行一次或多次产生相同的结果，是构建可靠系统的基石。

什么是幂等性

幂等操作：无论执行多少次，结果都相同

✅ 幂等操作示例：
- UPDATE users SET email='new@example.com' WHERE id=123
- DELETE FROM orders WHERE id=456
- PUT /users/123 (完整替换)

❌ 非幂等操作示例：
- INSERT INTO logs (message) VALUES ('event')  -- 每次插入新记录
- UPDATE accounts SET balance = balance + 100 WHERE id=123  -- 余额累加
- POST /orders (创建订单)  -- 每次创建新订单

HTTP方法幂等性

方法	幂等性	说明
GET	✅ 幂等	读取操作，无副作用
PUT	✅ 幂等	完整替换资源
DELETE	✅ 幂等	删除资源（删除已删除的资源仍返回成功）
POST	❌ 非幂等	创建资源，需要额外处理
PATCH	⚠️ 视情况	部分更新，取决于实现

幂等键模式（Idempotency Key）

实现原理

type IdempotencyHandler struct {
    store IdempotencyStore
    ttl   time.Duration
}

type IdempotencyStore interface {
    // 保存幂等键和响应
    Save(ctx context.Context, key string, response *IdempotentResponse) error
    
    // 查询幂等键
    Get(ctx context.Context, key string) (*IdempotentResponse, error)
    
    // 删除过期键
    Cleanup(ctx context.Context, before time.Time) error
}

type IdempotentResponse struct {
    StatusCode int               `json:"status_code"`
    Headers    map[string]string `json:"headers"`
    Body       []byte            `json:"body"`
    CreatedAt  time.Time         `json:"created_at"`
}

func (h *IdempotencyHandler) Middleware(next http.Handler) http.Handler {
    return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
        // 仅处理POST请求（非幂等方法）
        if r.Method != http.MethodPost {
            next.ServeHTTP(w, r)
            return
        }
        
        // 提取幂等键
        idempotencyKey := r.Header.Get("Idempotency-Key")
        if idempotencyKey == "" {
            http.Error(w, "Missing Idempotency-Key header", http.StatusBadRequest)
            return
        }
        
        // 检查是否已处理过
        cachedResponse, err := h.store.Get(r.Context(), idempotencyKey)
        if err == nil && cachedResponse != nil {
            // 返回缓存的响应
            h.writeResponse(w, cachedResponse)
            return
        }
        
        // 首次处理：记录响应
        recorder := &responseRecorder{
            ResponseWriter: w,
            body:           &bytes.Buffer{},
        }
        
        next.ServeHTTP(recorder, r)
        
        // 缓存响应
        response := &IdempotentResponse{
            StatusCode: recorder.statusCode,
            Headers:    recorder.Header(),
            Body:       recorder.body.Bytes(),
            CreatedAt:  time.Now(),
        }
        
        h.store.Save(r.Context(), idempotencyKey, response)
    })
}

func (h *IdempotencyHandler) writeResponse(w http.ResponseWriter, resp *IdempotentResponse) {
    for key, value := range resp.Headers {
        w.Header().Set(key, value)
    }
    w.Header().Set("X-Idempotent-Replay", "true")
    w.WriteHeader(resp.StatusCode)
    w.Write(resp.Body)
}

Redis幂等存储

type RedisIdempotencyStore struct {
    client *redis.Client
    ttl    time.Duration
}

func NewRedisIdempotencyStore(client *redis.Client, ttl time.Duration) *RedisIdempotencyStore {
    return &RedisIdempotencyStore{
        client: client,
        ttl:    ttl,
    }
}

func (s *RedisIdempotencyStore) Save(ctx context.Context, key string, response *IdempotentResponse) error {
    data, err := json.Marshal(response)
    if err != nil {
        return err
    }
    
    return s.client.Set(ctx, "idempotency:"+key, data, s.ttl).Err()
}

func (s *RedisIdempotencyStore) Get(ctx context.Context, key string) (*IdempotentResponse, error) {
    data, err := s.client.Get(ctx, "idempotency:"+key).Bytes()
    if err == redis.Nil {
        return nil, nil
    }
    if err != nil {
        return nil, err
    }
    
    var response IdempotentResponse
    err = json.Unmarshal(data, &response)
    return &response, err
}

客户端使用示例

// 支付客户端
type PaymentClient struct {
    httpClient *http.Client
}

func (c *PaymentClient) ProcessPayment(ctx context.Context, paymentID string, amount Money) error {
    // 生成幂等键（基于业务唯一标识）
    idempotencyKey := fmt.Sprintf("payment:%s", paymentID)
    
    reqBody := map[string]interface{}{
        "payment_id": paymentID,
        "amount":     amount.Amount,
        "currency":   amount.Currency,
    }
    
    bodyBytes, _ := json.Marshal(reqBody)
    req, _ := http.NewRequestWithContext(ctx, "POST", "https://api.example.com/payments", bytes.NewReader(bodyBytes))
    req.Header.Set("Content-Type", "application/json")
    req.Header.Set("Idempotency-Key", idempotencyKey)
    
    // 重试逻辑
    for attempt := 0; attempt < 3; attempt++ {
        resp, err := c.httpClient.Do(req)
        if err != nil {
            // 网络错误，重试（使用相同的幂等键）
            time.Sleep(time.Duration(attempt) * time.Second)
            continue
        }
        
        if resp.StatusCode == http.StatusOK || resp.StatusCode == http.StatusCreated {
            return nil
        }
        
        if resp.StatusCode == http.StatusTooManyRequests {
            // 限流，等待后重试
            time.Sleep(5 * time.Second)
            continue
        }
        
        // 其他错误，不重试
        return fmt.Errorf("payment failed: %d", resp.StatusCode)
    }
    
    return errors.New("payment failed after retries")
}

数据库唯一约束

防止重复创建

-- 订单表：使用唯一约束防止重复创建
CREATE TABLE orders (
    id UUID PRIMARY KEY,
    order_number VARCHAR(50) UNIQUE NOT NULL,  -- 唯一约束
    user_id UUID NOT NULL,
    total_amount DECIMAL(10,2) NOT NULL,
    status VARCHAR(20) NOT NULL,
    created_at TIMESTAMP DEFAULT NOW()
);

-- 支付记录表
CREATE TABLE payments (
    id UUID PRIMARY KEY,
    order_id UUID NOT NULL,
    payment_provider VARCHAR(50) NOT NULL,
    provider_transaction_id VARCHAR(100),
    amount DECIMAL(10,2) NOT NULL,
    status VARCHAR(20) NOT NULL,
    created_at TIMESTAMP DEFAULT NOW(),
    
    -- 同一订单同一提供商的交易ID唯一
    UNIQUE(order_id, payment_provider, provider_transaction_id)
);

type OrderService struct {
    db *sql.DB
}

func (s *OrderService) CreateOrder(ctx context.Context, cmd CreateOrderCommand) (*Order, error) {
    orderID := uuid.New()
    
    _, err := s.db.ExecContext(ctx, `
        INSERT INTO orders (id, order_number, user_id, total_amount, status)
        VALUES ($1, $2, $3, $4, 'created')
    `, orderID, cmd.OrderNumber, cmd.UserID, cmd.TotalAmount)
    
    if err != nil {
        if isUniqueViolation(err) {
            // 订单号已存在，返回现有订单
            return s.GetOrderByNumber(ctx, cmd.OrderNumber)
        }
        return nil, err
    }
    
    return &Order{
        ID:          orderID,
        OrderNumber: cmd.OrderNumber,
        UserID:      cmd.UserID,
        TotalAmount: cmd.TotalAmount,
        Status:      "created",
    }, nil
}

func isUniqueViolation(err error) bool {
    // PostgreSQL: 23505
    // MySQL: 1062
    return strings.Contains(err.Error(), "23505") || 
           strings.Contains(err.Error(), "1062")
}

状态机模式

防止重复状态转换

type OrderStateMachine struct {
    validTransitions map[OrderStatus][]OrderStatus
}

type OrderStatus string

const (
    StatusCreated   OrderStatus = "created"
    StatusPaid      OrderStatus = "paid"
    StatusShipped   OrderStatus = "shipped"
    StatusDelivered OrderStatus = "delivered"
    StatusCancelled OrderStatus = "cancelled"
)

func NewOrderStateMachine() *OrderStateMachine {
    return &OrderStateMachine{
        validTransitions: map[OrderStatus][]OrderStatus{
            StatusCreated:   {StatusPaid, StatusCancelled},
            StatusPaid:      {StatusShipped, StatusCancelled},
            StatusShipped:   {StatusDelivered},
            StatusDelivered: {},
            StatusCancelled: {},
        },
    }
}

func (sm *OrderStateMachine) CanTransition(from, to OrderStatus) bool {
    validTargets := sm.validTransitions[from]
    for _, target := range validTargets {
        if target == to {
            return true
        }
    }
    return false
}

type OrderService struct {
    db          *sql.DB
    stateMachine *OrderStateMachine
}

func (s *OrderService) MarkAsPaid(ctx context.Context, orderID string) error {
    // 使用乐观锁和状态检查
    result, err := s.db.ExecContext(ctx, `
        UPDATE orders 
        SET status = 'paid', updated_at = NOW()
        WHERE id = $1 AND status = 'created'
    `, orderID)
    
    if err != nil {
        return err
    }
    
    rowsAffected, _ := result.RowsAffected()
    if rowsAffected == 0 {
        // 检查订单状态
        var currentStatus string
        err := s.db.QueryRowContext(ctx, 
            "SELECT status FROM orders WHERE id = $1", orderID).Scan(&currentStatus)
        
        if err != nil {
            return err
        }
        
        if currentStatus == "paid" {
            // 已经是支付状态，幂等返回成功
            return nil
        }
        
        return fmt.Errorf("invalid state transition from %s to paid", currentStatus)
    }
    
    return nil
}

Token机制（Prevention Token）

两步提交

type OrderService struct {
    db    *sql.DB
    redis *redis.Client
}

// 步骤1：创建订单Token
func (s *OrderService) CreateOrderToken(ctx context.Context, userID string) (string, error) {
    token := uuid.New().String()
    
    // 保存Token信息（5分钟过期）
    tokenData := map[string]interface{}{
        "user_id":   userID,
        "created_at": time.Now().Unix(),
        "used":      false,
    }
    
    data, _ := json.Marshal(tokenData)
    err := s.redis.Set(ctx, "order_token:"+token, data, 5*time.Minute).Err()
    if err != nil {
        return "", err
    }
    
    return token, nil
}

// 步骤2：使用Token创建订单
func (s *OrderService) CreateOrder(ctx context.Context, token string, cmd CreateOrderCommand) (*Order, error) {
    // 验证Token
    tokenData, err := s.redis.Get(ctx, "order_token:"+token).Bytes()
    if err == redis.Nil {
        return nil, errors.New("invalid or expired token")
    }
    if err != nil {
        return nil, err
    }
    
    var data map[string]interface{}
    json.Unmarshal(tokenData, &data)
    
    if data["used"].(bool) {
        return nil, errors.New("token already used")
    }
    
    // 标记Token已使用（原子操作）
    success, err := s.redis.Eval(ctx, `
        local data = redis.call('GET', KEYS[1])
        local parsed = cjson.decode(data)
        if parsed.used == false then
            parsed.used = true
            redis.call('SET', KEYS[1], cjson.encode(parsed), 'EX', 300)
            return 1
        end
        return 0
    `, []string{"order_token:" + token}).Int()
    
    if err != nil {
        return nil, err
    }
    
    if success == 0 {
        return nil, errors.New("token already used")
    }
    
    // 创建订单
    return s.doCreateOrder(ctx, cmd)
}

重试策略

指数退避重试

type RetryConfig struct {
    MaxAttempts     int
    InitialInterval time.Duration
    MaxInterval     time.Duration
    Multiplier      float64
}

func RetryWithBackoff(ctx context.Context, config RetryConfig, operation func() error) error {
    var lastErr error
    interval := config.InitialInterval
    
    for attempt := 0; attempt < config.MaxAttempts; attempt++ {
        err := operation()
        if err == nil {
            return nil
        }
        
        lastErr = err
        
        // 检查是否应该重试
        if !shouldRetry(err) {
            return err
        }
        
        // 指数退避
        select {
        case <-time.After(interval):
            interval = time.Duration(float64(interval) * config.Multiplier)
            if interval > config.MaxInterval {
                interval = config.MaxInterval
            }
        case <-ctx.Done():
            return ctx.Err()
        }
    }
    
    return fmt.Errorf("max retries exceeded: %w", lastErr)
}

func shouldRetry(err error) bool {
    // 网络错误、超时、5xx错误可以重试
    if errors.Is(err, context.DeadlineExceeded) {
        return true
    }
    
    // HTTP状态码检查
    if httpErr, ok := err.(*HTTPError); ok {
        return httpErr.StatusCode >= 500 || httpErr.StatusCode == 429
    }
    
    return false
}

// 使用示例
err := RetryWithBackoff(ctx, RetryConfig{
    MaxAttempts:     5,
    InitialInterval: 1 * time.Second,
    MaxInterval:     30 * time.Second,
    Multiplier:      2.0,
}, func() error {
    return paymentClient.ProcessPayment(ctx, payment)
})

总结

幂等性设计模式选择指南：

模式	适用场景	优点	缺点
幂等键	API请求	通用性强	需要客户端配合
唯一约束	数据库操作	数据库级别保证	仅限数据库
状态机	状态转换	逻辑清晰	需要设计状态流
Token机制	表单提交	防止重复提交	增加复杂度

关键原则：

设计API时优先考虑幂等性
使用唯一标识符作为幂等键
合理设置幂等键的过期时间
结合重试策略处理网络故障
日志记录重复请求，便于排查

延伸阅读

继续阅读

探索更多技术文章

浏览归档，发现更多关于系统设计、工具链和工程实践的内容。

全部文章返回首页

引言