1、问题场景
以用户账户为例,如果允许同时对某个用户的账户进行修改的话,会导致某些修改被覆盖,使最后的结果不正确。
如:1.1、张三的账户中有100元。
1.2、张三的账户消费了50元。
1.3、张三的账户充值了100元。
我们希望的张三账户最终的结果是150元。如果1.2、1.3是并发执行的,按下面的方式执行的话,回事怎样的呢?
账户实体:
/** * 账户实体 * * @author caofanqi */ @Slf4j @Data @EqualsAndHashCode(callSuper = true) @Entity @Builder @Table(name = "jpa_account") @NoArgsConstructor @AllArgsConstructor public class Account extends AbstractID { /** * 简单代表一下账户所属人 */ private String accountName; @Column(columnDefinition = "DECIMAL(19, 2)") private BigDecimal balance; }
Repository接口:
/** * @author caofanqi */ public interface AccountRepository extends JpaRepositoryImplementation<Account,Long> { Account findByAccountName(String accountName); }
Service:
/** * * @author caofanqi */ @Service public class AccountServiceImpl implements AccountService { @Resource private AccountRepository accountRepository; @Override @Transactional(rollbackFor = Exception.class) public String addAccountMoney(String accountName, BigDecimal money){ System.out.println(Thread.currentThread().getName() + ",addAccountMoney start..."); try { TimeUnit.SECONDS.sleep(1); } catch (InterruptedException e) { e.printStackTrace(); } Account account = accountRepository.findByAccountName(accountName); System.out.println(Thread.currentThread().getName() + ",find balance : " + account.getBalance()); account.setBalance(account.getBalance().add(money)); try { TimeUnit.SECONDS.sleep(1); } catch (InterruptedException e) { e.printStackTrace(); } Account result = accountRepository.save(account); System.out.println(Thread.currentThread().getName() + ", update balance end ,balance : " + result.getBalance()); System.out.println(Thread.currentThread().getName() + ",addAccountMoney sleep..."); try { TimeUnit.SECONDS.sleep(5); } catch (InterruptedException e) { e.printStackTrace(); } System.out.println(Thread.currentThread().getName() + ",addAccountMoney end..."); return "success"; } }
数据库表中数据:
测试用例:
@Test void addAccountMoney() throws InterruptedException { CountDownLatch count = new CountDownLatch(2); ExecutorService executorService = Executors.newFixedThreadPool(2); executorService.execute(() -> { String result = accountService.addAccountMoney("张三的账户", BigDecimal.valueOf(-50)); System.out.println(Thread.currentThread().getName() + ",result : " + result); count.countDown(); }); TimeUnit.SECONDS.sleep(1); executorService.execute(() -> { String result = accountService.addAccountMoney("张三的账户", BigDecimal.valueOf(100)); System.out.println(Thread.currentThread().getName() + ",result : " + result); count.countDown(); }); count.await(10, TimeUnit.SECONDS); Account endAccount = accountRepository.findByAccountName("张三的账户"); System.out.println("final balance :" + endAccount.getBalance()); }
控制台打印及数据库结果:
这明显不是我们想要的正确答案,那怎么解决呢?这里提供几个方法,①如果是单JVM的话,可以使用Java的同步机制和Lock(估计这种情况很少见吧...)。②使用JPA为我们提供的乐观锁@Version。
③使用JPA为我们提供的@Lock中的悲观锁。
2、@Version
JPA提供的乐观锁,指定实体中的字段或属性作为乐观锁的version,该version用于确保并发操作的正确性。每个实体只能使用一个version属性或字段。version支持(int, Integer, short, Short, long, Long, java.sql.Timestamp)类型的属性或字段。
使用起来非常方便,我们只需要在实体中添加一个字段,并添加@Version注解就可以了。加了@Version后,insert和update的SQL语句都会带上version的操作。当乐观锁更新失败的时候,会抛出异常org.springframework.orm.ObjectOptimisticLockingFailureException。我们自己进行业务处理。
实体修改如下:
/** * 账户实体 * * @author caofanqi */ @Slf4j @Data @EqualsAndHashCode(callSuper = true) @Entity @Builder @Table(name = "jpa_account") @NoArgsConstructor @AllArgsConstructor public class Account extends AbstractID { /** * 简单代表一下账户所属人 */ private String accountName; @Column(columnDefinition = "DECIMAL(19, 2)") private BigDecimal balance; /** * 乐观锁version */ @Version private Integer version; }
重新插入一条数据,可以看到数据库中如下
修改Service方法如下:
@Override @Transactional(rollbackFor = Exception.class) public String addAccountMoney(String accountName, BigDecimal money){ try { updateAccount(accountName, money); return "success"; }catch (ObjectOptimisticLockingFailureException e){ //记录日志,重新操作... return "fail"; } } @Transactional(propagation = Propagation.REQUIRES_NEW,rollbackFor = Exception.class) public void updateAccount(String accountName, BigDecimal money) { System.out.println(Thread.currentThread().getName() + ",addAccountMoney start..."); try { TimeUnit.SECONDS.sleep(1); } catch (InterruptedException e) { e.printStackTrace(); } Account account = accountRepository.findByAccountName(accountName); System.out.println(Thread.currentThread().getName() + ",find balance : " + account.getBalance()); account.setBalance(account.getBalance().add(money)); try { TimeUnit.SECONDS.sleep(1); } catch (InterruptedException e) { e.printStackTrace(); } Account result = accountRepository.save(account); System.out.println(Thread.currentThread().getName() + ", update balance end ,balance : " + result.getBalance()); System.out.println(Thread.currentThread().getName() + ",addAccountMoney sleep..."); try { TimeUnit.SECONDS.sleep(5); } catch (InterruptedException e) { e.printStackTrace(); } System.out.println(Thread.currentThread().getName() + ",addAccountMoney end..."); }
重新运行测试用例:
这样只有和我们上次版本一样的时候才会更新,就不会出现互相覆盖的问题,保证了数据的原子性。但是如果我们的业务就是需要让两次都必须成功,那么可以使用下面的悲观锁来实现。
3、@Lock
spring-data-jpa为我们提供了@Lock注解,指定查询方法要使用的锁定模式。可以添加在派生查询上,也可以重写父类CRUD的方法,添加该注解。@Lock只有一个value属性,为LockModeType枚举类型,我们主要看以下里面的悲观锁PESSIMISTIC_WRITE。
修改Repository如下:
/** * @author caofanqi */ public interface AccountRepository extends JpaRepositoryImplementation<Account,Long> { @Lock(LockModeType.PESSIMISTIC_WRITE) Account findByAccountName(String accountName); }
恢复数据库表数据为100,并将@Version注解去掉,运行测试用例控制台打印如下:
pool-1-thread-1,addAccountMoney start... pool-1-thread-2,addAccountMoney start... Hibernate: select account0_.id as id1_0_, account0_.account_name as account_2_0_, account0_.balance as balance3_0_, account0_.version as version4_0_ from cfq_jpa_account account0_ where account0_.account_name=? for update pool-1-thread-1,find balance : 100.00 Hibernate: select account0_.id as id1_0_, account0_.account_name as account_2_0_, account0_.balance as balance3_0_, account0_.version as version4_0_ from cfq_jpa_account account0_ where account0_.account_name=? for update pool-1-thread-1, update balance end ,balance : 50.00 pool-1-thread-1,addAccountMoney sleep... pool-1-thread-1,addAccountMoney end... Hibernate: update cfq_jpa_account set account_name=?, balance=?, version=? where id=? pool-1-thread-2,find balance : 50.00 pool-1-thread-1,result : success pool-1-thread-2, update balance end ,balance : 150.00 pool-1-thread-2,addAccountMoney sleep... Hibernate: select account0_.id as id1_0_, account0_.account_name as account_2_0_, account0_.balance as balance3_0_, account0_.version as version4_0_ from cfq_jpa_account account0_ where account0_.account_name=? for update pool-1-thread-2,addAccountMoney end... Hibernate: update cfq_jpa_account set account_name=?, balance=?, version=? where id=? final balance :150.00 2019-12-08 17:20:43.915 INFO 4160 --- [ main] o.s.t.c.transaction.TransactionContext : Committed transaction for test: [DefaultTestContext@7674f035 testClass = AccountServiceImplTest, testInstance = cn.caofanqi.study.studyspringdatajpa.service.impl.AccountServiceImplTest@46d69ca4, testMethod = addAccountMoney@AccountServiceImplTest, testException = [null], mergedContextConfiguration = [WebMergedContextConfiguration@69e153c5 testClass = AccountServiceImplTest, locations = '{}', classes = '{class cn.caofanqi.study.studyspringdatajpa.StudySpringDataJpaApplication}', contextInitializerClasses = '[]', activeProfiles = '{}', propertySourceLocations = '{}', propertySourceProperties = '{org.springframework.boot.test.context.SpringBootTestContextBootstrapper=true}', contextCustomizers = set[org.springframework.boot.test.context.filter.ExcludeFilterContextCustomizer@9353778, org.springframework.boot.test.json.DuplicateJsonObjectContextCustomizerFactory$DuplicateJsonObjectContextCustomizer@1700915, org.springframework.boot.test.mock.mockito.MockitoContextCustomizer@0, org.springframework.boot.test.web.client.TestRestTemplateContextCustomizer@31c88ec8, org.springframework.boot.test.autoconfigure.properties.PropertyMappingContextCustomizer@0, org.springframework.boot.test.autoconfigure.web.servlet.WebDriverContextCustomizerFactory$Customizer@20ce78ec], resourceBasePath = 'src/main/webapp', contextLoader = 'org.springframework.boot.test.context.SpringBootContextLoader', parent = [null]], attributes = map['org.springframework.test.context.web.ServletTestExecutionListener.activateListener' -> true, 'org.springframework.test.context.web.ServletTestExecutionListener.populatedRequestContextHolder' -> true, 'org.springframework.test.context.web.ServletTestExecutionListener.resetRequestContextHolder' -> true]]
可以看到查询语句通过for update进行加锁。得到了我们想要的150结果。
注意:for update ,如果不通过索引条件检索数据,那么InnoDB将对表中的所有记录加锁,实际效果跟锁表一样。我们进行测试,在数据库中在添加一条记录,如下:
执行下面测试用例:
/** * for update ,如果不通过索引条件检索数据,那么InnoDB将对表中的所有记录加锁,实际效果跟锁表一样 */ @Test void addAccountMoney2() throws InterruptedException { CountDownLatch count = new CountDownLatch(2); ExecutorService executorService = Executors.newFixedThreadPool(2); executorService.execute(() -> { String result = accountService.addAccountMoney("张三的账户", BigDecimal.valueOf(-50)); System.out.println(Thread.currentThread().getName() + ",result : " + result); count.countDown(); }); TimeUnit.SECONDS.sleep(1); executorService.execute(() -> { String result = accountService.addAccountMoney("李四的账户", BigDecimal.valueOf(100)); System.out.println(Thread.currentThread().getName() + ",result : " + result); count.countDown(); }); count.await(20, TimeUnit.SECONDS); }
控制台打印结果:
可以看到并不是并行进行的更新,我们就该实体类,重新生成数据库表,并插入数据(或直接修改数据库)
/** * 账户实体 * * @author caofanqi */ @Slf4j @Data @EqualsAndHashCode(callSuper = true) @Entity @Builder @Table(name = "jpa_account") @NoArgsConstructor @AllArgsConstructor public class Account extends AbstractID { /** * 简单代表一下账户所属人 */ @Column(unique = true,nullable = false) private String accountName; @Column(columnDefinition = "DECIMAL(19, 2)") private BigDecimal balance; /** * 乐观锁version */ // @Version private Integer version; }
重新运行测试用例:
我们在使用的过程中要根据自己的业务进行选择。
参考连接:https://blog.csdn.net/u014316026/article/details/78726459
https://blog.csdn.net/loophome/article/details/79867174
源码地址:https://github.com/caofanqi/study-spring-data-jpa