Spring data Jpa,Mybatis,读写锁,@Lock 使用

Posted 2020-10-31 高因咖啡

Spring data jpa 支持注解式的读写锁(悲观锁),实际上这个东西硬编码也简单,但是基于Jpa 命名方式定义的Sql,只能用注解添加支持读写锁了,

不了解读写锁的可以点这里 

mysql读写锁及事务

并且推荐

PESSIMISTIC_READ,
PESSIMISTIC_WRITE,
而不是

READ,
WRITE,
但是官方文档貌似没有更新这个案例,踩了一些坑.

新建一个实体Book.java

/**
 * User: laizhenwei
 * Date: 2018-04-18 Time: 9:04
 * Description:
 */
@Entity
@Table(name = "test_book")
@Alias("Book")
@Getter
@Setter
@NoArgsConstructor
@AllArgsConstructor
public class Book extends AbstractJbatisIdEntity{

    private static final long serialVersionUID = -1L;

    private String name;

    private String author;

}

BookRepository.java

/**
 * User: laizhenwei
 * Date: 2018-04-18 Time: 9:11
 * Description:
 */
public interface BookRepository extends JpaRepository<Book,String> {
    @Lock(LockModeType.PESSIMISTIC_READ)
    Book findTop1ByName(String name);
}

BookServiceImpl TimeUnit.SECONDS.sleep(20); 是为了让事务延迟提交,好测试save操作需要阻塞到读写释放才能提交

    public static final  CountDownLatch readCount = new CountDownLatch(1);
    public static final  CountDownLatch saveCount = new CountDownLatch(1);

    @Transactional(propagation = Propagation.REQUIRES_NEW)
    @Override
    public Book save(Book book){
        Book book1 = null;
        try {
            readCount.await();
            book1 = getRepository().save(book);
            saveCount.countDown();
        } catch (InterruptedException e) {
            e.printStackTrace();
        }
        return book1;
    }

    @Transactional(propagation = Propagation.REQUIRES_NEW)
    @Override
    public Book findTop1ByName(String name){
        Book book = getRepository().findTop1ByName(name);
        try {
            readCount.countDown();
            TimeUnit.SECONDS.sleep(20);
        } catch (InterruptedException e) {
            e.printStackTrace();
        }
        return book;
    }

JunitTest 先添加一条数据,待会要锁这个数据

    @Test
    public void save(){
        Book book = new Book();
        book.setName("百年孤独");
        book.setAuthor("加西亚·马尔克斯");
        bookRepository.save(book);
    }

开启两条线程,一条先加上读锁,然后睡眠一会,另一条线程去修改这个对象的时候,需要阻塞到读事务提交以后才会成功

第二个查询动作不会阻塞,因为读锁只对写操作限制(这里用直接用bookRepository,是为了避免CountDownLatch 再一次阻塞而已)

    @Test
    @Transactional
    public void findByName() throws InterruptedException {

        new Thread(()->bookService.findTop1ByName("百年孤独")).start();
        BookServiceImpl.readCount.await();
        Book book =bookRepository.findTop1ByName("百年孤独");

        book.setAuthor("加西亚·马尔克斯5");
        new Thread(()->bookService.save(book)).start();
        BookServiceImpl.saveCount.await();

    }

有个有趣的现象,如果直接运行第二次,会发现不用阻塞,就能save成功,因为数据并没有做任何修改.

再注释掉@Lock跑一次,修改 book.setAuthor("加西亚·马尔克斯5");再保存也不需要等待.

 

Mybatis下的实现,就是手动编码而已

/**
 * User: laizhenwei
 * Date: 2018-04-18 Time: 9:12
 */
@Mapper
public interface BookMapper extends BaseMapper<Book> {

    @Select("select * from test_book where name=#{name} limit 1 lock in share mode")
    Book findTop1ByName(String name);

}

Service  TimeUnit.SECONDS.sleep(20); 是为了让事务延迟提交,好测试save操作需要阻塞到读写释放才能提交

    @Transactional(propagation = Propagation.REQUIRES_NEW)
    @Override
    public Book mapperFindTop1ByName(String name){
        Book book = getMapper().findTop1ByName(name);
        try {
            readCount.countDown();
            TimeUnit.SECONDS.sleep(20);
        } catch (InterruptedException e) {
            e.printStackTrace();
        }
        return book;
    }

JunitTest (这里用直接用bookRepository,是为了避免CountDownLatch 再一次阻塞而已)

    @Test
    @Transactional
    public void findByName() throws InterruptedException {

        new Thread(()->bookService.mapperFindTop1ByName("百年孤独")).start();
        BookServiceImpl.readCount.await();
        Book book =bookRepository.findTop1ByName("百年孤独"); 

        book.setAuthor("加西亚·马尔克斯3");
        new Thread(()->bookService.save(book)).start();
        BookServiceImpl.saveCount.await();

    }

测试效果与Jpa一样.