interview/questions/01-分布式系统/数据库锁机制.md

数据库锁机制面试指南

1. 行锁、表锁、页锁

行锁（Row Lock）

定义：锁定数据库表的某一行或多行数据

特点：

• 锁粒度最细，并发性能最好
• 开销大，需要更多的锁资源
• 适合高并发场景

InnoDB 行锁类型：

1. 记录锁：锁定单行记录
2. 间隙锁：锁定间隙，防止幻读
3. 临键锁：记录锁+间隙锁的组合

代码示例：

-- 记录锁
SELECT * FROM user WHERE id = 1 FOR UPDATE;

-- 间隙锁
SELECT * FROM user WHERE id BETWEEN 1 AND 10 FOR UPDATE;

-- 临键锁
SELECT * FROM user WHERE id > 5 FOR UPDATE;

Java 实现：

// 行锁实现示例
public class RowLockExample {

    @Autowired
    private JdbcTemplate jdbcTemplate;

    // 记录锁
    public User getUserWithRecordLock(Long userId) {
        String sql = "SELECT * FROM user WHERE id = ? FOR UPDATE";
        return jdbcTemplate.queryForObject(sql, new Object[]{userId}, (rs, rowNum) -> {
            User user = new User();
            user.setId(rs.getLong("id"));
            user.setName(rs.getString("name"));
            user.setAge(rs.getInt("age"));
            return user;
        });
    }

    // 间隙锁
    public List<User> getUsersWithGapLock(String minAge, String maxAge) {
        String sql = "SELECT * FROM user WHERE age BETWEEN ? AND ? FOR UPDATE";
        return jdbcTemplate.query(sql, new Object[]{minAge, maxAge}, (rs, rowNum) -> {
            User user = new User();
            user.setId(rs.getLong("id"));
            user.setName(rs.getString("name"));
            user.setAge(rs.getInt("age"));
            return user;
        });
    }
}

表锁（Table Lock）

定义：锁定整个表，所有操作都需要获取表级锁

特点：

• 锁粒度最粗，并发性能最差
• 开销小，实现简单
• 适用于读多写少或操作简单的场景

代码示例：

-- 表锁
LOCK TABLES user WRITE;  -- 写锁，阻塞其他所有操作
-- 执行操作
UNLOCK TABLES;          -- 释放锁

-- 读锁
LOCK TABLES user READ;   -- 读锁，允许并发读，阻塞写
-- 执行读操作
UNLOCK TABLES;          -- 释放锁

Java 实现：

// 表锁实现示例
public class TableLockExample {

    @Autowired
    private DataSource dataSource;

    // 写锁
    public void executeWithWriteLock(Runnable operation) {
        try (Connection conn = dataSource.getConnection()) {
            // 获取写锁
            conn.createStatement().execute("LOCK TABLES user WRITE");

            try {
                // 执行操作
                operation.run();
            } finally {
                // 释放锁
                conn.createStatement().execute("UNLOCK TABLES");
            }
        } catch (SQLException e) {
            throw new RuntimeException("执行写锁操作失败", e);
        }
    }

    // 读锁
    public List<User> executeWithReadLock(Runnable operation) {
        try (Connection conn = dataSource.getConnection()) {
            // 获取读锁
            conn.createStatement().execute("LOCK TABLES user READ");

            try {
                // 执行读操作
                operation.run();
            } finally {
                // 释放锁
                conn.createStatement().execute("UNLOCK TABLES");
            }
        } catch (SQLException e) {
            throw new RuntimeException("执行读锁操作失败", e);
        }
        return Collections.emptyList();
    }
}

页锁（Page Lock）

定义：锁定数据库表的数据页，介于行锁和表锁之间

特点：

• 锁粒度居中
• 开销居中
• 适用于中等到高并发场景

InnoDB 页锁：

• InnoDB 默认使用行锁，页锁主要用于某些特定操作
• 在全表扫描时可能会升级为表锁

代码示例：

-- InnoDB 自动管理页锁
-- 通常不需要手动控制
-- 全表扫描时可能触发页锁
SELECT * FROM user WHERE age > 100 FOR UPDATE;

锁粒度对比

锁类型	锁粒度	并发性能	开销	适用场景
表锁	最粗	最差	最小	简单操作、维护
页锁	中等	中等	中等	中等并发
行锁	最细	最好	最大	高并发、高吞吐

2. 共享锁、排他锁

共享锁（Shared Lock / S Lock）

定义：多个事务可以同时持有共享锁，但不能持有排他锁

特点：

• 读锁，允许多个事务同时读取
• 不允许修改数据
• 用于读多写少的场景

代码示例：

-- 共享锁语法
SELECT * FROM user WHERE id = 1 LOCK IN SHARE MODE;

-- Java 实现
public class SharedLockExample {

    @Autowired
    private JdbcTemplate jdbcTemplate;

    public User getUserWithSharedLock(Long userId) {
        String sql = "SELECT * FROM user WHERE id = ? LOCK IN SHARE MODE";
        return jdbcTemplate.queryForObject(sql, new Object[]{userId}, (rs, rowNum) -> {
            User user = new User();
            user.setId(rs.getLong("id"));
            user.setName(rs.getString("name"));
            user.setAge(rs.getInt("age"));
            return user;
        });
    }

    public List<User> getUsersWithSharedLock() {
        String sql = "SELECT * FROM user WHERE age > 18 LOCK IN SHARE MODE";
        return jdbcTemplate.query(sql, (rs, rowNum) -> {
            User user = new User();
            user.setId(rs.getLong("id"));
            user.setName(rs.getString("name"));
            user.setAge(rs.getInt("age"));
            return user;
        });
    }
}

排他锁（Exclusive Lock / X Lock）

定义：只有一个事务可以持有排他锁，阻止其他事务获取任何锁

特点：

• 写锁，独占访问
• 阻止其他事务读取和修改
• 用于写操作

代码示例：

-- 排他锁语法
SELECT * FROM user WHERE id = 1 FOR UPDATE;

-- Java 实现
public class ExclusiveLockExample {

    @Autowired
    private JdbcTemplate jdbcTemplate;

    public User getUserWithExclusiveLock(Long userId) {
        String sql = "SELECT * FROM user WHERE id = ? FOR UPDATE";
        return jdbcTemplate.queryForObject(sql, new Object[]{userId}, (rs, rowNum) -> {
            User user = new User();
            user.setId(rs.getLong("id"));
            user.setName(rs.getString("name"));
            user.setAge(rs.getInt("age"));
            return user;
        });
    }

    public void updateUserWithExclusiveLock(Long userId, String newName) {
        String sql = "UPDATE user SET name = ? WHERE id = ?";
        jdbcTemplate.update(sql, newName, userId);
    }
}

锁兼容性矩阵

已有锁	请求共享锁	请求排他锁
共享锁	兼容	不兼容
排他锁	不兼容	不兼容

兼容性图解：

锁兼容性矩阵
          S锁    X锁
    S锁   ✓      ✗
    X锁   ✗      ✗

锁升级与降级

// 锁升级示例
public class LockUpgradeExample {

    @Autowired
    private JdbcTemplate jdbcTemplate;

    // 锁升级：从共享锁升级到排他锁
    public void upgradeLock(Long userId) {
        try (Connection conn = jdbcTemplate.getDataSource().getConnection()) {
            // 1. 获取共享锁
            conn.setAutoCommit(false);
            User user = getUserWithSharedLock(userId);

            // 2. 检查是否需要升级
            if (needUpgrade(user)) {
                // 3. 升级为排他锁
                try {
                    conn.createStatement().execute(
                        "SELECT * FROM user WHERE id = ? FOR UPDATE");

                    // 4. 执行更新操作
                    updateUserWithExclusiveLock(userId, "New Name");

                    conn.commit();
                } catch (SQLException e) {
                    conn.rollback();
                    throw new RuntimeException("锁升级失败", e);
                }
            } else {
                conn.commit();
            }
        } catch (SQLException e) {
            throw new RuntimeException("锁操作失败", e);
        }
    }

    // 锁降级示例
    public void downgradeLock(Long userId) {
        try (Connection conn = jdbcTemplate.getDataSource().getConnection()) {
            conn.setAutoCommit(false);

            // 1. 获取排他锁
            User user = getUserWithExclusiveLock(userId);

            // 2. 降级为共享锁
            try {
                conn.createStatement().execute(
                    "SELECT * FROM user WHERE id = ? LOCK IN SHARE MODE");

                // 3. 执行读操作
                List<User> users = getUsersWithSharedLock();

                conn.commit();
            } catch (SQLException e) {
                conn.rollback();
                throw new RuntimeException("锁降级失败", e);
            }
        } catch (SQLException e) {
            throw new RuntimeException("锁操作失败", e);
        }
    }
}

3. 意向锁、间隙锁、临键锁

意向锁（Intention Lock）

定义：表级锁，表示事务意图获取行级锁

类型：

1. 意向共享锁（IS）：事务意图获取某些行的共享锁
2. 意向排他锁（IX）：事务意图获取某些行的排他锁

特点：

• 提高锁效率，避免逐行检查
• 表级锁与行级锁协同工作

代码示例：

-- 意向锁（InnoDB 自动管理）
-- 不需要手动设置
-- 在获取行锁时自动添加

-- 查看意向锁
SELECT * FROM information_schema.innodb_locks;

Java 实现：

// 意向锁示例
public class IntentionLockExample {

    @Autowired
    private JdbcTemplate jdbcTemplate;

    // 多个事务同时获取行锁，意向锁提高效率
    public void multipleRowLocks() {
        ExecutorService executor = Executors.newFixedThreadPool(5);

        for (int i = 0; i < 5; i++) {
            final int userId = i + 1;
            executor.submit(() -> {
                try (Connection conn = jdbcTemplate.getDataSource().getConnection()) {
                    conn.setAutoCommit(false);

                    // 获取行锁（自动添加意向锁）
                    User user = getUserWithExclusiveLock((long) userId);

                    // 处理数据
                    Thread.sleep(1000);

                    conn.commit();
                } catch (Exception e) {
                    throw new RuntimeException("行锁操作失败", e);
                }
            });
        }

        executor.shutdown();
    }
}

间隙锁（Gap Lock）

定义：锁定索引记录之间的间隙，防止其他事务插入数据

特点：

• 防止幻读
• 只在 REPEATABLE READ 隔离级别下有效
• 使用间隙锁锁定范围查询

代码示例：

-- 间隙锁示例
-- 用户表数据：id: 1, 5, 10, 15, 20

-- 锁定id=5和id=10之间的间隙
SELECT * FROM user WHERE id > 5 AND id < 10 FOR UPDATE;
-- 锁定间隙：(5, 10)

-- 锁定id>10的间隙
SELECT * FROM user WHERE id > 10 FOR UPDATE;
-- 锁定间隙：(10, ∞)

-- 防止插入
-- 其他事务无法在(5, 10)之间插入新记录

Java 实现：

// 间隙锁示例
public class GapLockExample {

    @Autowired
    private JdbcTemplate jdbcTemplate;

    // 间隙锁防止幻读
    public void preventPhantomRead() {
        try (Connection conn = jdbcTemplate.getDataSource().getConnection()) {
            conn.setAutoCommit(false);

            // 查询并锁定间隙
            List<User> users = jdbcTemplate.query(
                "SELECT * FROM user WHERE age BETWEEN 20 AND 30 FOR UPDATE",
                (rs, rowNum) -> {
                    User user = new User();
                    user.setId(rs.getLong("id"));
                    user.setName(rs.getString("name"));
                    user.setAge(rs.getInt("age"));
                    return user;
                });

            // 处理数据
            System.out.println("找到 " + users.size() + " 个用户");

            // 模拟长时间处理
            Thread.sleep(5000);

            conn.commit();

            // 其他事务无法在age 20-30之间插入新记录
        } catch (Exception e) {
            throw new RuntimeException("间隙锁操作失败", e);
        }
    }

    // 优化后的范围查询（避免间隙锁）
    public void optimizedRangeQuery() {
        try (Connection conn = jdbcTemplate.getDataSource().getConnection()) {
            conn.setAutoCommit(false);

            // 使用主键查询避免间隙锁
            List<User> users = jdbcTemplate.query(
                "SELECT * FROM user WHERE age >= 20 AND age <= 30 FOR UPDATE",
                (rs, rowNum) -> {
                    User user = new User();
                    user.setId(rs.getLong("id"));
                    user.setName(rs.getString("name"));
                    user.setAge(rs.getInt("age"));
                    return user;
                });

            conn.commit();
        } catch (Exception e) {
            throw new RuntimeException("范围查询失败", e);
        }
    }
}

临键锁（Next-Key Lock）

定义：记录锁与间隙锁的组合，锁定记录本身及其相邻间隙

特点：

• 解决幻读问题
• 在 REPEATABLE READ 隔离级别下默认使用
• 锁定记录+前一个间隙

代码示例：

-- 临键锁示例
-- 用户表数据：id: 1, 5, 10, 15, 20

-- 锁定id=10及其前后间隙
SELECT * FROM user WHERE id = 10 FOR UPDATE;
-- 锁定：(5, 10] + (10, 15)

-- 锁定范围查询
SELECT * FROM user WHERE id <= 10 FOR UPDATE;
-- 锁定：(-∞, 1] + (1, 5] + (5, 10] + (10, 15)

Java 实现：

// 临键锁示例
public class NextKeyLockExample {

    @Autowired
    private JdbcTemplate jdbcTemplate;

    // 临键锁防止幻读
    public void preventPhantomReadWithNextKeyLock() {
        try (Connection conn = jdbcTemplate.getDataSource().getConnection()) {
            conn.setAutoCommit(false);

            // 查询并获取临键锁
            List<User> users = jdbcTemplate.query(
                "SELECT * FROM user WHERE age <= 30 FOR UPDATE",
                (rs, rowNum) -> {
                    User user = new User();
                    user.setId(rs.getLong("id"));
                    user.setName(rs.getString("name"));
                    user.setAge(rs.getInt("age"));
                    return user;
                });

            // 处理数据
            System.out.println("找到 " + users.size() + " 个用户");

            // 防止其他事务插入age<=30的记录
            conn.commit();
        } catch (Exception e) {
            throw new RuntimeException("临键锁操作失败", e);
        }
    }

    // 优化查询避免不必要的锁
    public void optimizedQueryWithNextKeyLock() {
        try (Connection conn = jdbcTemplate.getDataSource().getConnection()) {
            conn.setAutoCommit(false);

            // 使用主键查询减少锁范围
            User user = jdbcTemplate.queryForObject(
                "SELECT * FROM user WHERE id = ? FOR UPDATE",
                new Object[]{10L},
                (rs, rowNum) -> {
                    User u = new User();
                    u.setId(rs.getLong("id"));
                    u.setName(rs.getString("name"));
                    u.setAge(rs.getInt("age"));
                    return u;
                });

            conn.commit();
        } catch (Exception e) {
            throw new RuntimeException("查询失败", e);
        }
    }
}

4. 乐观锁、悲观锁

乐观锁（Optimistic Locking）

定义：假设冲突不常发生，只在更新时检查冲突

实现方式：

1. 版本号：使用版本号字段
2. 时间戳：使用更新时间
3. CAS：Compare And Swap

代码示例：

// 乐观锁实现
@Component
public class OptimisticLockService {

    @Autowired
    private UserRepository userRepository;

    // 使用版本号实现乐观锁
    @Transactional
    public User updateUserWithVersion(Long userId, String newName) {
        User user = userRepository.findById(userId)
            .orElseThrow(() -> new RuntimeException("用户不存在"));

        // 尝试更新
        int affected = userRepository.updateWithVersion(
            userId, newName, user.getVersion());

        if (affected == 0) {
            throw new OptimisticLockException("数据已被其他事务修改");
        }

        return userRepository.findById(userId).orElse(null);
    }

    // 使用时间戳实现乐观锁
    @Transactional
    public User updateUserWithTimestamp(Long userId, String newName) {
        User user = userRepository.findById(userId)
            .orElseThrow(() -> new RuntimeException("用户不存在"));

        // 尝试更新
        int affected = userRepository.updateWithTimestamp(
            userId, newName, user.getUpdateTime());

        if (affected == 0) {
            throw new OptimisticLockException("数据已被其他事务修改");
        }

        return userRepository.findById(userId).orElse(null);
    }
}

// Repository 实现
@Repository
public class UserRepository {

    @Autowired
    private JdbcTemplate jdbcTemplate;

    // 带版本号的更新
    public int updateWithVersion(Long userId, String newName, Integer version) {
        String sql = "UPDATE user SET name = ?, version = version + 1 " +
                    "WHERE id = ? AND version = ?";
        return jdbcTemplate.update(sql, newName, userId, version);
    }

    // 带时间戳的更新
    public int updateWithTimestamp(Long userId, String newName, Timestamp timestamp) {
        String sql = "UPDATE user SET name = ?, update_time = NOW() " +
                    "WHERE id = ? AND update_time = ?";
        return jdbcTemplate.update(sql, newName, userId, timestamp);
    }
}

悲观锁（Pessimistic Locking）

定义：假设冲突经常发生，直接加锁

实现方式：

1. 行锁：SELECT ... FOR UPDATE
2. 表锁：LOCK TABLES
3. 间隙锁：防止幻读

代码示例：

// 悲观锁实现
@Component
public class PessimisticLockService {

    @Autowired
    private UserRepository userRepository;

    // 使用行锁实现悲观锁
    @Transactional
    public User updateUserWithRowLock(Long userId, String newName) {
        // 1. 加锁
        User user = userRepository.findByIdWithLock(userId)
            .orElseThrow(() -> new RuntimeException("用户不存在"));

        // 2. 更新数据
        user.setName(newName);
        user.setUpdateTime(new Date());

        // 3. 保存（锁在事务提交时释放）
        return userRepository.save(user);
    }

    // 使用表锁实现悲观锁
    @Transactional
    public void updateUserWithTableLock(Long userId, String newName) {
        // 1. 获取表锁
        jdbcTemplate.update("LOCK TABLES user WRITE");

        try {
            // 2. 执行更新
            int affected = userRepository.update(userId, newName);
            if (affected == 0) {
                throw new RuntimeException("更新失败");
            }

            // 3. 提交事务时自动释放锁
        } finally {
            // 确保释放锁
            jdbcTemplate.update("UNLOCK TABLES");
        }
    }
}

// Repository 实现
@Repository
public class UserRepository {

    @Autowired
    private JdbcTemplate jdbcTemplate;

    // 带锁的查询
    public Optional<User> findByIdWithLock(Long userId) {
        String sql = "SELECT * FROM user WHERE id = ? FOR UPDATE";
        return jdbcTemplate.queryForObject(sql, new Object[]{userId},
            (rs, rowNum) -> {
                User user = new User();
                user.setId(rs.getLong("id"));
                user.setName(rs.getString("name"));
                user.setAge(rs.getInt("age"));
                user.setVersion(rs.getInt("version"));
                user.setUpdateTime(rs.getTimestamp("update_time"));
                return user;
            });
    }

    // 普通更新
    public int update(Long userId, String newName) {
        String sql = "UPDATE user SET name = ? WHERE id = ?";
        return jdbcTemplate.update(sql, newName, userId);
    }
}

乐观锁 vs 悲观锁

特性	乐观锁	悲观锁
冲突假设	冲突少	冲突多
性能	读性能好，写有冲突时开销大	写性能好，读性能差
适用场景	读多写少，冲突少	写多，冲突多
实现复杂度	简单	复杂
锁类型	无锁，版本号控制	行锁、表锁等

优化策略

// 锁策略优化
@Service
public class LockOptimizationService {

    @Autowired
    private UserRepository userRepository;

    // 根据业务场景选择锁策略
    public User updateUser(Long userId, String newName) {
        // 1. 首先使用乐观锁
        try {
            return updateUserWithVersion(userId, newName);
        } catch (OptimisticLockException e) {
            // 2. 乐观锁冲突，使用悲观锁重试
            return updateUserWithPessimisticLock(userId, newName);
        }
    }

    // 乐观锁重试机制
    @Retryable(value = OptimisticLockException.class, maxAttempts = 3, backoff = @Backoff(delay = 100))
    public User updateUserWithVersion(Long userId, String newName) {
        User user = userRepository.findById(userId)
            .orElseThrow(() -> new RuntimeException("用户不存在"));

        int affected = userRepository.updateWithVersion(
            userId, newName, user.getVersion());

        if (affected == 0) {
            throw new OptimisticLockException("版本冲突");
        }

        return userRepository.findById(userId).orElse(null);
    }

    // 悲观锁实现
    @Transactional
    public User updateUserWithPessimisticLock(Long userId, String newName) {
        User user = userRepository.findByIdWithLock(userId)
            .orElseThrow(() -> new RuntimeException("用户不存在"));

        user.setName(newName);
        return userRepository.save(user);
    }
}

5. 死锁的产生和解决

死锁的产生条件

四个必要条件：

1. 互斥条件：资源不能共享
2. 请求和保持：进程保持资源的同时请求其他资源
3. 不可剥夺：资源不能被强制释放
4. 循环等待：形成等待环路

死锁示例：

// 死锁示例
public class DeadlockExample {

    @Autowired
    private JdbcTemplate jdbcTemplate;

    public void createDeadlock() {
        // 线程1：获取资源A，等待资源B
        new Thread(() -> {
            try (Connection conn1 = jdbcTemplate.getDataSource().getConnection()) {
                conn1.setAutoCommit(false);

                // 获取用户表的锁
                conn1.createStatement().execute("SELECT * FROM user WHERE id = 1 FOR UPDATE");
                System.out.println("线程1获取了用户表锁");

                // 等待订单表锁
                Thread.sleep(1000);

                // 尝试获取订单表锁
                conn1.createStatement().execute("SELECT * FROM order WHERE id = 1 FOR UPDATE");
                System.out.println("线程1获取了订单表锁");

                conn1.commit();
            } catch (Exception e) {
                System.out.println("线程1异常: " + e.getMessage());
            }
        }).start();

        // 线程2：获取资源B，等待资源A
        new Thread(() -> {
            try (Connection conn2 = jdbcTemplate.getDataSource().getConnection()) {
                conn2.setAutoCommit(false);

                // 获取订单表的锁
                conn2.createStatement().execute("SELECT * FROM order WHERE id = 1 FOR UPDATE");
                System.out.println("线程2获取了订单表锁");

                // 等待用户表锁
                Thread.sleep(1000);

                // 尝试获取用户表锁
                conn2.createStatement().execute("SELECT * FROM user WHERE id = 1 FOR UPDATE");
                System.out.println("线程2获取了用户表锁");

                conn2.commit();
            } catch (Exception e) {
                System.out.println("线程2异常: " + e.getMessage());
            }
        }).start();
    }
}

死锁检测和监控

1. MySQL 死锁检测：

-- 查看死锁信息
SHOW ENGINE INNODB STATUS;

-- 查看锁等待
SELECT * FROM information_schema.innodb_lock_waits;

2. 死锁监控实现：

// 死锁监控
@Component
public class DeadlockMonitor {

    @Autowired
    private JdbcTemplate jdbcTemplate;

    @Scheduled(fixedRate = 60000) // 每分钟监控一次
    public void monitorDeadlocks() {
        String sql = "SELECT * FROM information_schema.innodb_lock_waits";

        try {
            List<Map<String, Object>> deadlocks = jdbcTemplate.queryForList(sql);

            if (!deadlocks.isEmpty()) {
                // 发送告警
                notificationService.sendAlert("检测到死锁", deadlocks);

                // 记录日志
                log.error("检测到死锁: {}", deadlocks);
            }
        } catch (Exception e) {
            log.error("死锁监控失败", e);
        }
    }

    // 死锁分析
    public void analyzeDeadlock(String deadlockReport) {
        // 解析死锁报告
        DeadlockAnalysis analysis = parseDeadlockReport(deadlockReport);

        // 生成解决方案建议
        List<String> suggestions = generateSuggestions(analysis);

        // 记录分析结果
        log.info("死锁分析结果: {}", suggestions);
    }
}

死锁预防策略

1. 锁排序：

// 锁排序预防死锁
public class LockOrderService {

    private final Map<String, Object> lockMap = new ConcurrentHashMap<>();

    // 使用固定的锁顺序
    @Transactional
    public void transferMoney(String fromAccount, String toAccount, BigDecimal amount) {
        // 1. 按账户ID排序，确保锁顺序一致
        String account1 = fromAccount.compareTo(toAccount) < 0 ? fromAccount : toAccount;
        String account2 = fromAccount.compareTo(toAccount) < 0 ? toAccount : fromAccount;

        // 2. 按固定顺序加锁
        synchronized (getLock(account1)) {
            synchronized (getLock(account2)) {
                // 执行转账
                transfer(fromAccount, toAccount, amount);
            }
        }
    }

    private Object getLock(String account) {
        return lockMap.computeIfAbsent(account, k -> new Object());
    }
}

2. 超时机制：

// 超时机制预防死锁
public class TimeoutLockService {

    @Autowired
    private DataSource dataSource;

    @Transactional(timeout = 3) // 事务超时3秒
    public void executeWithTimeout(Runnable operation) {
        try {
            operation.run();
        } catch (TransactionTimedOutException e) {
            // 超时回滚，释放锁
            throw new RuntimeException("操作超时，已回滚", e);
        }
    }

    // 设置锁等待超时
    public void executeWithLockTimeout(Runnable operation, long timeout) {
        try (Connection conn = dataSource.getConnection()) {
            conn.setAutoCommit(false);

            // 设置锁等待超时
            conn.createStatement().execute(
                "SET innodb_lock_wait_timeout = " + timeout);

            try {
                operation.run();
                conn.commit();
            } catch (SQLException e) {
                conn.rollback();
                throw new RuntimeException("锁等待超时", e);
            }
        } catch (SQLException e) {
            throw new RuntimeException("数据库操作失败", e);
        }
    }
}

3. 避免嵌套锁：

// 避免嵌套锁示例
public class NestedLockService {

    // 不好的示例：嵌套锁
    @Transactional
    public void badNestedLock(Long userId, Long orderId) {
        // 锁1
        User user = userRepository.findByIdWithLock(userId);

        // 业务逻辑
        processUser(user);

        // 锁2（嵌套锁可能导致死锁）
        Order order = orderRepository.findByIdWithLock(orderId);

        processOrder(order);
    }

    // 好的示例：统一锁
    @Transactional
    public void goodSingleLock(Long userId, Long orderId) {
        // 统一获取需要的锁
        List<User> users = userRepository.findUsersByIdsWithLock(Arrays.asList(userId));
        List<Order> orders = orderRepository.findOrdersByIdsWithLock(Arrays.asList(orderId));

        // 处理业务逻辑
        processUsersAndOrders(users, orders);
    }
}

死锁恢复策略

1. 自动重试：

// 自动重试机制
@Service
public class RetryService {

    @Retryable(value = DeadlockLoserDataAccessException.class,
               maxAttempts = 3,
               backoff = @Backoff(delay = 100))
    public <T> T executeWithRetry(Callable<T> operation) {
        return operation.call();
    }

    // 使用示例
    @Transactional
    public User updateUserWithRetry(Long userId, String newName) {
        return executeWithRetry(() -> {
            User user = userRepository.findByIdWithLock(userId);
            user.setName(newName);
            return userRepository.save(user);
        });
    }
}

2. 降级处理：

// 降级处理策略
@Service
public class FallbackService {

    @Autowired
    private CacheManager cacheManager;

    @Transactional
    public User updateUserWithFallback(Long userId, String newName) {
        try {
            // 正常更新
            User user = userRepository.findByIdWithLock(userId);
            user.setName(newName);
            return userRepository.save(user);
        } catch (Exception e) {
            // 降级处理：先更新缓存
            updateCache(userId, newName);

            // 异步更新数据库
            asyncUpdateDatabase(userId, newName);

            // 返回缓存中的数据
            return getCachedUser(userId);
        }
    }

    private void updateCache(Long userId, String newName) {
        Cache cache = cacheManager.getCache("userCache");
        User user = new User();
        user.setId(userId);
        user.setName(newName);
        cache.put(userId, user);
    }

    private User getCachedUser(Long userId) {
        Cache cache = cacheManager.getCache("userCache");
        return cache.get(userId, User.class);
    }
}

6. 实际项目中的锁使用

电商系统锁策略

1. 库存管理：

// 库存管理锁策略
@Service
public class InventoryService {

    @Autowired
    private InventoryRepository inventoryRepository;

    // 乐观锁实现库存扣减
    @Transactional
    public boolean reduceStockWithOptimisticLock(Long productId, int quantity) {
        Inventory inventory = inventoryRepository.findByProductId(productId)
            .orElseThrow(() -> new RuntimeException("商品不存在"));

        // 检查库存
        if (inventory.getStock() < quantity) {
            return false;
        }

        // 乐观锁更新
        int affected = inventoryRepository.reduceStockWithVersion(
            productId, quantity, inventory.getVersion());

        if (affected == 0) {
            // 版本冲突，重试或抛出异常
            throw new OptimisticLockException("库存已被其他订单占用");
        }

        return true;
    }

    // 悲观锁实现库存扣减
    @Transactional
    public boolean reduceStockWithPessimisticLock(Long productId, int quantity) {
        // 悲观锁防止并发扣减
        Inventory inventory = inventoryRepository.findByProductIdWithLock(productId)
            .orElseThrow(() -> new RuntimeException("商品不存在"));

        if (inventory.getStock() < quantity) {
            return false;
        }

        inventory.setStock(inventory.getStock() - quantity);
        inventoryRepository.save(inventory);

        return true;
    }

    // 批量库存扣减
    @Transactional
    public boolean batchReduceStock(List<OrderItem> items) {
        // 使用事务隔离级别避免死锁
        try {
            for (OrderItem item : items) {
                boolean success = reduceStockWithOptimisticLock(
                    item.getProductId(), item.getQuantity());
                if (!success) {
                    // 回滚整个事务
                    throw new RuntimeException("库存不足");
                }
            }
            return true;
        } catch (Exception e) {
            throw new RuntimeException("批量扣减库存失败", e);
        }
    }
}

2. 订单处理：

// 订单处理锁策略
@Service
public class OrderService {

    @Autowired
    private OrderRepository orderRepository;

    @Autowired
    private PaymentService paymentService;

    // 订单创建锁
    @Transactional(isolation = Isolation.REPEATABLE_READ)
    public Order createOrder(OrderDTO orderDTO) {
        // 1. 检查用户和商品状态
        User user = userRepository.findById(orderDTO.getUserId())
            .orElseThrow(() -> new RuntimeException("用户不存在"));

        List<Product> products = productRepository.findByIds(
            orderDTO.getItems().stream()
                .map(OrderItemDTO::getProductId)
                .collect(Collectors.toList()));

        // 2. 创建订单
        Order order = new Order();
        BeanUtils.copyProperties(orderDTO, order);
        order.setStatus("PENDING");
        order.setCreateTime(new Date());

        order = orderRepository.save(order);

        // 3. 扣减库存（悲观锁）
        for (OrderItem item : order.getItems()) {
            boolean success = inventoryService.reduceStockWithPessimisticLock(
                item.getProductId(), item.getQuantity());
            if (!success) {
                throw new RuntimeException("库存不足");
            }
        }

        // 4. 预扣金额（乐观锁）
        try {
            paymentService.reserveAmount(order.getUserId(), order.getTotalAmount());
        } catch (Exception e) {
            // 回滚库存
            rollbackInventory(order.getItems());
            throw e;
        }

        return order;
    }

    // 订单取消锁
    @Transactional
    public void cancelOrder(Long orderId) {
        Order order = orderRepository.findByIdWithLock(orderId)
            .orElseThrow(() -> new RuntimeException("订单不存在"));

        // 检查订单状态
        if (!"PENDING".equals(order.getStatus())) {
            throw new RuntimeException("订单状态不允许取消");
        }

        // 更新订单状态
        order.setStatus("CANCELLED");
        order.setUpdateTime(new Date());
        orderRepository.save(order);

        // 释放库存
        for (OrderItem item : order.getItems()) {
            inventoryService.releaseStock(item.getProductId(), item.getQuantity());
        }

        // 释放金额
        paymentService.releaseAmount(order.getUserId(), order.getTotalAmount());
    }
}

社交系统锁策略

1. 点赞功能：

// 点赞功能锁策略
@Service
public class LikeService {

    @Autowired
    private LikeRepository likeRepository;

    @Autowired
    private PostRepository postRepository;

    // 乐观锁实现点赞
    @Transactional
    public LikeDTO addLike(Long postId, Long userId) {
        // 1. 检查是否已经点赞
        boolean alreadyLiked = likeRepository.existsByPostIdAndUserId(postId, userId);
        if (alreadyLiked) {
            throw new RuntimeException("已经点赞");
        }

        // 2. 创建点赞记录（乐观锁）
        Like like = new Like();
        like.setPostId(postId);
        like.setUserId(userId);
        like.setCreateTime(new Date());

        like = likeRepository.save(like);

        // 3. 更新点赞数（乐观锁）
        updateLikeCount(postId, 1);

        return convertToDTO(like);
    }

    // 批量点赞
    @Transactional
    public void batchAddLikes(List<Long> postIds, Long userId) {
        for (Long postId : postIds) {
            addLike(postId, userId);
        }
    }

    // 分布式锁实现点赞
    @Transactional
    public LikeDTO addLikeWithDistributedLock(Long postId, Long userId) {
        String lockKey = "like_" + postId + "_" + userId;

        // 获取分布式锁
        try {
            boolean locked = redisLock.tryLock(lockKey, 5, TimeUnit.SECONDS);
            if (!locked) {
                throw new RuntimeException("系统繁忙，请稍后重试");
            }

            // 检查是否已经点赞
            boolean alreadyLiked = likeRepository.existsByPostIdAndUserId(postId, userId);
            if (alreadyLiked) {
                throw new RuntimeException("已经点赞");
            }

            // 创建点赞记录
            Like like = new Like();
            like.setPostId(postId);
            like.setUserId(userId);
            like.setCreateTime(new Date());

            like = likeRepository.save(like);

            // 更新点赞数
            updateLikeCount(postId, 1);

            return convertToDTO(like);

        } finally {
            // 释放锁
            redisLock.unlock(lockKey);
        }
    }
}

2. 评论系统：

// 评论系统锁策略
@Service
public class CommentService {

    @Autowired
    private CommentRepository commentRepository;

    // 间隙锁防止评论重复
    @Transactional(isolation = Isolation.REPEATABLE_READ)
    public CommentDTO addComment(CommentDTO commentDTO) {
        // 1. 获取帖子间隙锁
        Post post = postRepository.findByIdWithGapLock(commentDTO.getPostId())
            .orElseThrow(() -> new RuntimeException("帖子不存在"));

        // 2. 创建评论
        Comment comment = new Comment();
        BeanUtils.copyProperties(commentDTO, comment);
        comment.setCreateTime(new Date());

        comment = commentRepository.save(comment);

        // 3. 更新帖子统计
        post.setCommentCount(post.getCommentCount() + 1);
        postRepository.save(post);

        return convertToDTO(comment);
    }

    // 评论回复锁
    @Transactional
    public CommentDTO replyComment(CommentDTO replyDTO) {
        // 1. 获取父评论锁
        Comment parentComment = commentRepository.findByIdWithLock(replyDTO.getParentId())
            .orElseThrow(() -> new RuntimeException("父评论不存在"));

        // 2. 创建回复
        replyDTO.setParentId(parentComment.getId());
        replyDTO.setPostId(parentComment.getPostId());

        Comment reply = new Comment();
        BeanUtils.copyProperties(replyDTO, reply);
        reply.setCreateTime(new Date());

        reply = commentRepository.save(reply);

        // 3. 更新统计
        parentComment.setReplyCount(parentComment.getReplyCount() + 1);
        commentRepository.save(parentComment);

        return convertToDTO(reply);
    }
}

金融系统锁策略

1. 转账服务：

// 转账服务锁策略
@Service
public class TransferService {

    @Autowired
    private AccountRepository accountRepository;

    // 分布式锁实现转账
    @Transactional(isolation = Isolation.SERIALIZABLE)
    public void transferWithDistributedLock(Long fromAccountId, Long toAccountId, BigDecimal amount) {
        String lockKey = "transfer_" + Math.min(fromAccountId, toAccountId) + "_" +
                        Math.max(fromAccountId, toAccountId);

        try {
            // 获取分布式锁
            boolean locked = redisLock.tryLock(lockKey, 10, TimeUnit.SECONDS);
            if (!locked) {
                throw new RuntimeException("转账繁忙，请稍后重试");
            }

            transfer(fromAccountId, toAccountId, amount);

        } finally {
            // 释放锁
            redisLock.unlock(lockKey);
        }
    }

    // 乐观锁实现转账
    @Transactional
    public void transferWithOptimisticLock(Long fromAccountId, Long toAccountId, BigDecimal amount) {
        // 1. 获取账户信息（带乐观锁）
        Account fromAccount = accountRepository.findByIdWithVersion(fromAccountId)
            .orElseThrow(() -> new RuntimeException("转出账户不存在"));

        Account toAccount = accountRepository.findByIdWithVersion(toAccountId)
            .orElseThrow(() -> new RuntimeException("转入账户不存在"));

        // 2. 检查余额
        if (fromAccount.getBalance().compareTo(amount) < 0) {
            throw new RuntimeException("余额不足");
        }

        // 3. 执行转账（乐观锁）
        boolean success = transferWithVersion(fromAccount, toAccount, amount);
        if (!success) {
            throw new RuntimeException("转账失败，可能余额已变动");
        }
    }

    private boolean transferWithVersion(Account fromAccount, Account toAccount, BigDecimal amount) {
        try {
            // 扣减转出账户
            fromAccount.setBalance(fromAccount.getBalance().subtract(amount));

            // 增加转入账户
            toAccount.setBalance(toAccount.getBalance().add(amount));

            // 批量更新（乐观锁）
            accountRepository.saveAll(Arrays.asList(fromAccount, toAccount));

            return true;
        } catch (OptimisticLockException e) {
            // 版本冲突，重试
            return false;
        }
    }
}

2. 余额查询：

// 余额查询锁策略
@Service
public class BalanceQueryService {

    @Autowired
    private AccountRepository accountRepository;

    // 乐观锁实现余额查询
    @Transactional(readOnly = true)
    public BigDecimal getBalance(Long accountId) {
        Account account = accountRepository.findById(accountId)
            .orElseThrow(() -> new RuntimeException("账户不存在"));

        // 使用缓存优化
        BigDecimal balance = balanceCache.get(accountId);
        if (balance == null) {
            balance = account.getBalance();
            balanceCache.put(accountId, balance);
        }

        return balance;
    }

    // 悲观锁实现余额查询
    @Transactional(readOnly = true)
    public BigDecimal getBalanceWithLock(Long accountId) {
        Account account = accountRepository.findByIdWithLock(accountId)
            .orElseThrow(() -> new RuntimeException("账户不存在"));

        return account.getBalance();
    }
}

7. 阿里 P7 加分项

分布式锁实现

1. Redis 分布式锁：

// Redis 分布式锁实现
@Component
public class RedisLock {

    @Autowired
    private RedisTemplate<String, String> redisTemplate;

    private static final String LOCK_PREFIX = "lock:";
    private static final long DEFAULT_EXPIRE_TIME = 30; // 秒

    /**
     * 获取分布式锁
     */
    public boolean tryLock(String lockKey, long waitTime, TimeUnit timeUnit) {
        String key = LOCK_PREFIX + lockKey;
        long expireTime = DEFAULT_EXPIRE_TIME;

        try {
            long startTime = System.currentTimeMillis();
            long waitMillis = timeUnit.toMillis(waitTime);

            while (true) {
                // 尝试获取锁
                Boolean success = redisTemplate.opsForValue().setIfAbsent(
                    key,
                    Thread.currentThread().getName(),
                    expireTime,
                    TimeUnit.SECONDS);

                if (Boolean.TRUE.equals(success)) {
                    return true;
                }

                // 检查是否超时
                if (System.currentTimeMillis() - startTime > waitMillis) {
                    return false;
                }

                // 避免CPU空转
                Thread.sleep(100);
            }
        } catch (InterruptedException e) {
            Thread.currentThread().interrupt();
            return false;
        }
    }

    /**
     * 释放分布式锁
     */
    public void unlock(String lockKey) {
        String key = LOCK_PREFIX + lockKey;
        redisTemplate.delete(key);
    }

    /**
     * 可重入分布式锁
     */
    private final ConcurrentHashMap<String, Integer> lockCountMap = new ConcurrentHashMap<>();

    public boolean tryReentrantLock(String lockKey, long waitTime, TimeUnit timeUnit) {
        Thread currentThread = Thread.currentThread();
        String key = LOCK_PREFIX + lockKey;

        // 检查是否是当前线程持有锁
        Integer count = lockCountMap.get(key);
        if (count != null && count > 0) {
            lockCountMap.put(key, count + 1);
            return true;
        }

        // 获取新锁
        if (tryLock(lockKey, waitTime, timeUnit)) {
            lockCountMap.put(key, 1);
            return true;
        }

        return false;
    }

    public void releaseReentrantLock(String lockKey) {
        String key = LOCK_PREFIX + lockKey;
        Integer count = lockCountMap.get(key);

        if (count == null) {
            return;
        }

        if (count > 1) {
            lockCountMap.put(key, count - 1);
        } else {
            lockCountMap.remove(key);
            unlock(lockKey);
        }
    }
}

// Redis 分布式锁使用示例
@Service
public class RedisLockService {

    @Autowired
    private RedisLock redisLock;

    @Autowired
    private InventoryService inventoryService;

    public boolean reduceStockWithRedisLock(Long productId, int quantity) {
        String lockKey = "inventory_" + productId;

        try {
            // 获取分布式锁
            boolean locked = redisLock.tryLock(lockKey, 5, TimeUnit.SECONDS);
            if (!locked) {
                return false;
            }

            // 扣减库存
            return inventoryService.reduceStock(productId, quantity);

        } finally {
            // 释放锁
            redisLock.unlock(lockKey);
        }
    }
}

2. Zookeeper 分布式锁：

// Zookeeper 分布式锁实现
@Component
public class ZookeeperLock {

    @Autowired
    private CuratorFramework curatorFramework;

    /**
     * 创建分布式锁
     */
    public InterProcessMutex createDistributedLock(String lockPath) {
        return new InterProcessMutex(curatorFramework, lockPath);
    }

    /**
     * 获取锁
     */
    public boolean tryLock(String lockPath, long timeout, TimeUnit unit) {
        InterProcessMutex lock = createDistributedLock(lockPath);

        try {
            return lock.acquire(timeout, unit);
        } catch (Exception e) {
            throw new RuntimeException("获取分布式锁失败", e);
        }
    }

    /**
     * 释放锁
     */
    public void releaseLock(String lockPath) {
        InterProcessMutex lock = createDistributedLock(lockPath);

        try {
            lock.release();
        } catch (Exception e) {
            throw new RuntimeException("释放分布式锁失败", e);
        }
    }
}

// Zookeeper 分布式锁使用示例
@Service
public class ZookeeperLockService {

    @Autowired
    private ZookeeperLock zookeeperLock;

    @Autowired
    private PaymentService paymentService;

    public boolean processPaymentWithZkLock(Long orderId, BigDecimal amount) {
        String lockPath = "/payment/lock/" + orderId;

        try {
            // 获取分布式锁
            boolean locked = zookeeperLock.tryLock(lockPath, 10, TimeUnit.SECONDS);
            if (!locked) {
                return false;
            }

            // 处理支付
            return paymentService.processPayment(orderId, amount);

        } finally {
            // 释放锁
            zookeeperLock.releaseLock(lockPath);
        }
    }
}

高级锁优化策略

1. 自适应锁策略：

// 自适应锁策略
@Service
public class AdaptiveLockService {

    @Autowired
    private JdbcTemplate jdbcTemplate;

    @Autowired
    private RedisLock redisLock;

    // 根据业务复杂度选择锁策略
    public <T> T executeWithAdaptiveLock(Callable<T> operation, String lockKey, BusinessType type) {
        switch (type) {
            case SIMPLE:
                return executeWithoutLock(operation);
            case NORMAL:
                return executeWithOptimisticLock(operation);
            case COMPLEX:
                return executeWithPessimisticLock(operation, lockKey);
            case CRITICAL:
                return executeWithDistributedLock(operation, lockKey);
            default:
                return executeWithOptimisticLock(operation);
        }
    }

    private <T> T executeWithoutLock(Callable<T> operation) {
        try {
            return operation.call();
        } catch (Exception e) {
            throw new RuntimeException("操作失败", e);
        }
    }

    private <T> T executeWithOptimisticLock(Callable<T> operation) {
        int retryCount = 0;
        final int maxRetries = 3;

        while (retryCount < maxRetries) {
            try {
                return operation.call();
            } catch (OptimisticLockException e) {
                retryCount++;
                if (retryCount >= maxRetries) {
                    throw new RuntimeException("乐观锁重试次数已用尽", e);
                }
                // 指数退避
                try {
                    Thread.sleep(100 * (1 << retryCount));
                } catch (InterruptedException ie) {
                    Thread.currentThread().interrupt();
                    throw new RuntimeException("线程被中断", ie);
                }
            }
        }
        throw new RuntimeException("操作失败");
    }

    private <T> T executeWithPessimisticLock(Callable<T> operation, String lockKey) {
        try (Connection conn = jdbcTemplate.getDataSource().getConnection()) {
            conn.setAutoCommit(false);

            // 获取悲观锁
            conn.createStatement().execute(
                "SELECT * FROM business_data WHERE key = '" + lockKey + "' FOR UPDATE");

            try {
                T result = operation.call();
                conn.commit();
                return result;
            } catch (Exception e) {
                conn.rollback();
                throw e;
            }
        } catch (SQLException e) {
            throw new RuntimeException("悲观锁操作失败", e);
        }
    }

    private <T> T executeWithDistributedLock(Callable<T> operation, String lockKey) {
        try {
            boolean locked = redisLock.tryLock(lockKey, 5, TimeUnit.SECONDS);
            if (!locked) {
                throw new RuntimeException("获取分布式锁失败");
            }

            try {
                return operation.call();
            } finally {
                redisLock.unlock(lockKey);
            }
        } catch (Exception e) {
            throw new RuntimeException("分布式锁操作失败", e);
        }
    }
}

// 业务类型枚举
public enum BusinessType {
    SIMPLE,    // 简单操作，不需要锁
    NORMAL,    // 普通操作，使用乐观锁
    COMPLEX,   // 复杂操作，使用悲观锁
    CRITICAL   // 关键操作，使用分布式锁
}

2. 智能锁监控：

// 智能锁监控
@Component
public class SmartLockMonitor {

    @Autowired
    private MeterRegistry meterRegistry;

    @Autowired
    private JdbcTemplate jdbcTemplate;

    // 监控锁使用情况
    @Scheduled(fixedRate = 60000)
    public void monitorLockUsage() {
        // 1. 监控数据库锁等待
        monitorDatabaseLocks();

        // 2. 监控分布式锁等待
        monitorDistributedLocks();

        // 3. 监控锁超时
        monitorLockTimeout();
    }

    private void monitorDatabaseLocks() {
        String sql = "SELECT COUNT(*) FROM information_schema.innodb_lock_waits";

        Integer waitCount = jdbcTemplate.queryForObject(sql, Integer.class);

        if (waitCount != null && waitCount > 0) {
            meterRegistry.gauge("database.lock.waits", waitCount);

            // 发送告警
            if (waitCount > 10) {
                notificationService.sendAlert("数据库锁等待过多", waitCount);
            }
        }
    }

    private void monitorDistributedLocks() {
        // 监控 Redis 锁等待
        Set<String> keys = redisTemplate.keys("lock:*");
        if (keys != null) {
            meterRegistry.gauge("distributed.lock.count", keys.size());

            // 监控锁等待时间
            for (String key : keys) {
                String value = redisTemplate.opsForValue().get(key);
                if (value != null) {
                    long ttl = redisTemplate.getExpire(key);
                    if (ttl > 0) {
                        meterRegistry.gauge("distributed.lock.ttl", ttl, k -> ttl);
                    }
                }
            }
        }
    }

    private void monitorLockTimeout() {
        // 监控锁超时
        String sql = "SELECT COUNT(*) FROM transaction WHERE status = 'timeout'";

        Integer timeoutCount = jdbcTemplate.queryForObject(sql, Integer.class);

        if (timeoutCount != null && timeoutCount > 0) {
            meterRegistry.gauge("transaction.lock.timeout", timeoutCount);
        }
    }

    // 智能锁推荐
    public LockRecommendation recommendLockStrategy(String tableName, String operation) {
        // 1. 分析表特征
        TableFeature feature = analyzeTableFeature(tableName);

        // 2. 分析操作频率
        OperationFrequency frequency = analyzeOperationFrequency(tableName, operation);

        // 3. 基于规则推荐锁策略
        if (feature.isHighWrite() && frequency.isHighFrequency()) {
            return new LockRecommendation(
                LockType.OPTIMISTIC,
                "乐观锁适合高写入频率场景",
                3 // 重试次数
            );
        } else if (feature.isCritical() && frequency.isLowFrequency()) {
            return new LockRecommendation(
                LockType.PESSIMISTIC,
                "悲观锁适合关键数据",
                1
            );
        } else {
            return new LockRecommendation(
                LockType.ADAPTIVE,
                "自适应锁策略",
                3
            );
        }
    }
}

总结

数据库锁机制是并发控制的核心，需要根据业务场景选择合适的锁策略：

1. 掌握各种锁类型：行锁、表锁、页锁、共享锁、排他锁、意向锁、间隙锁、临键锁
2. 理解锁的兼容性：锁与锁之间的兼容关系
3. 掌握乐观锁和悲观锁：两种锁策略的优缺点和适用场景
4. 避免死锁：通过锁排序、超时、避免嵌套锁等方式预防死锁
5. 实际项目应用：电商、社交、金融等不同场景的锁策略
6. 高级特性：分布式锁、智能锁监控、自适应锁策略

在实际项目中，需要平衡性能和一致性，选择合适的锁策略，并持续优化锁的使用。