MYSQL批量插入并发场景下的DEADLOCK
一、背景公元2023-10-12(周四)上午，组内的亚梅反馈，用户生成标签报死锁异常
二、排查异常日志
查到当时报错的日志
具体异常信息如下
server-provider-info-2023-10-12.0.log:2023-10-12 09:40:50.593 INFOcom.emax.user.user.provider.UserTagAPIImpl:? - 签约完成，更新用户标签信息异常，用户idorg.apache.ibatis.exceptions.PersistenceException:
server-provider-info-2023-10-12.0.log-### Error flushing statements.Cause: org.apache.ibatis.executor.BatchExecutorException: com.emax.user.user.mapper.UserTagMapper.insert (batch index #1) failed. Cause: java.sql.BatchUpdateException: Deadlock found when trying to get lock; try restarting transaction
server-provider-info-2023-10-12.0.log-### Cause: org.apache.ibatis.executor.BatchExecutorException: com.emax.user.user.mapper.UserTagMapper.insert (batch index #1)failed. Cause: java.sql.BatchUpdateException: Deadlock found when trying to get lock; try restarting transaction
server-provider-info-2023-10-12.0.log-at org.apache.ibatis.exceptions.ExceptionFactory.wrapException(ExceptionFactory.java:30)
server-provider-info-2023-10-12.0.log-at org.apache.ibatis.session.defaults.DefaultSqlSession.flushStatements(DefaultSqlSession.java:254)
server-provider-info-2023-10-12.0.log-at com.baomidou.mybatisplus.extension.service.impl.ServiceImpl.saveBatch(ServiceImpl.java:127)
server-provider-info-2023-10-12.0.log-at com.baomidou.mybatisplus.extension.service.IService.saveBatch(IService.java:58)
server-provider-info-2023-10-12.0.log-at com.baomidou.mybatisplus.extension.service.IService$$FastClassBySpringCGLIB$$f8525d18.invoke(<generated>)
server-provider-info-2023-10-12.0.log-at org.springframework.cglib.proxy.MethodProxy.invoke(MethodProxy.java:218)
server-provider-info-2023-10-12.0.log-at org.springframework.aop.framework.CglibAopProxy$CglibMethodInvocation.invokeJoinpoint(CglibAopProxy.java:749)
server-provider-info-2023-10-12.0.log-at org.springframework.aop.framework.ReflectiveMethodInvocation.proceed(ReflectiveMethodInvocation.java:163)
server-provider-info-2023-10-12.0.log-at org.springframework.transaction.interceptor.TransactionAspectSupport.invokeWithinTransaction(TransactionAspectSupport.java:294)
server-provider-info-2023-10-12.0.log-at org.springframework.transaction.interceptor.TransactionInterceptor.invoke(TransactionInterceptor.java:98)
server-provider-info-2023-10-12.0.log-at org.springframework.aop.framework.ReflectiveMethodInvocation.proceed(ReflectiveMethodInvocation.java:186)
server-provider-info-2023-10-12.0.log-at org.springframework.aop.framework.CglibAopProxy$DynamicAdvisedInterceptor.intercept(CglibAopProxy.java:688)
server-provider-info-2023-10-12.0.log-at com.emax.user.user.service.UserTagManager$$EnhancerBySpringCGLIB$$67ca0067.saveBatch(<generated>)
server-provider-info-2023-10-12.0.log-at com.emax.user.user.provider.UserTagAPIImpl.addUserTagByUser(UserTagAPIImpl.java:236)
server-provider-info-2023-10-12.0.log-at com.emax.user.user.provider.UserTagAPIImpl.addTagAfterSign(UserTagAPIImpl.java:116)
server-provider-info-2023-10-12.0.log-at com.emax.user.user.provider.UserTagAPIImpl$$FastClassBySpringCGLIB$$2547ce26.invoke(<generated>)
server-provider-info-2023-10-12.0.log-at org.springframework.cglib.proxy.MethodProxy.invoke(MethodProxy.java:218)
server-provider-info-2023-10-12.0.log-at org.springframework.aop.framework.CglibAopProxy$CglibMethodInvocation.invokeJoinpoint(CglibAopProxy.java:749)
--
server-provider-info-2023-10-12.0.log-at io.seata.rm.datasource.StatementProxy.lambda$executeBatch$9(StatementProxy.java:129)
server-provider-info-2023-10-12.0.log-at io.seata.rm.datasource.exec.ExecuteTemplate.execute(ExecuteTemplate.java:73)
server-provider-info-2023-10-12.0.log-at io.seata.rm.datasource.exec.ExecuteTemplate.execute(ExecuteTemplate.java:51)
server-provider-info-2023-10-12.0.log-at io.seata.rm.datasource.StatementProxy.executeBatch(StatementProxy.java:129)
server-provider-info-2023-10-12.0.log-at sun.reflect.GeneratedMethodAccessor1346.invoke(Unknown Source)
server-provider-info-2023-10-12.0.log-at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
server-provider-info-2023-10-12.0.log-at java.lang.reflect.Method.invoke(Method.java:498)
server-provider-info-2023-10-12.0.log-at org.apache.ibatis.logging.jdbc.PreparedStatementLogger.invoke(PreparedStatementLogger.java:78)
server-provider-info-2023-10-12.0.log-at com.sun.proxy.$Proxy501.executeBatch(Unknown Source)
server-provider-info-2023-10-12.0.log-at com.baomidou.mybatisplus.core.executor.MybatisBatchExecutor.doFlushStatements(MybatisBatchExecutor.java:132)
server-provider-info-2023-10-12.0.log-... 71 common frames omitted
server-provider-info-2023-10-12.0.log-Caused by: com.mysql.cj.jdbc.exceptions.MySQLTransactionRollbackException: Deadlock found when trying to get lock; try restarting transaction
server-provider-info-2023-10-12.0.log-at com.mysql.cj.jdbc.exceptions.SQLError.createSQLException(SQLError.java:123)
server-provider-info-2023-10-12.0.log-at com.mysql.cj.jdbc.exceptions.SQLError.createSQLException(SQLError.java:97)
server-provider-info-2023-10-12.0.log-at com.mysql.cj.jdbc.exceptions.SQLExceptionsMapping.translateException(SQLExceptionsMapping.java:122)
server-provider-info-2023-10-12.0.log-at com.mysql.cj.jdbc.ClientPreparedStatement.executeInternal(ClientPreparedStatement.java:970)
server-provider-info-2023-10-12.0.log-at com.mysql.cj.jdbc.ClientPreparedStatement.executeUpdateInternal(ClientPreparedStatement.java:1109)
server-provider-info-2023-10-12.0.log-at com.mysql.cj.jdbc.ClientPreparedStatement.executeBatchSerially(ClientPreparedStatement.java:849)
server-provider-info-2023-10-12.0.log-... 89 common frames omitted
server-provider-info-2023-10-12.0.log-
三、代码所在地
查看发生Deadlock的代码现场。
代码如下：
/**
* 增加用户标签
* @param user 用户信息
* @param tagNameList 标签集合
*/
private void addUserTagByUser(User user, List<String> tagNameList){
log.info("用户添加标签入参信息，当前用户id[{}]，标签信息[{}]", user.getUserId(), JSON.toJSONString(tagNameList));
Set<String> existTags = new HashSet<>();
final List<UserTag> userTagList = userTagManager.listUserTagByUserId(user.getUserId());
if(CollectionUtils.isNotEmpty(userTagList)){
existTags = userTagList.stream().map(UserTag::getTagName).collect(Collectors.toSet());
}
//遍历合集 如果不存在旧数据 直接都新增 存在 判断当前数据是否在旧数据中
List<UserTag> newInsertList = new ArrayList<>();
for(String name : tagNameList){
if (StringUtils.isNotBlank(name) && existTags.add(name)) {
newInsertList.add(userTagManager.buildEntity(user.getUserId(), user.getIdcardNo(), name));
}
}
userTagManager.saveBatch(newInsertList);
}
观察代码，大致逻辑是，先根据用户id查询出所有标签信息，计算出来需要新增的用户标签，批量入库。其中，user_tag表存储用户的标签数据。两个关键字段 user_id（用户id）和tag_name（标签）上面具有唯一索引，插入的数据也主要是这两个字段。
从这段代码并不容易判断出来如何会产生死锁。推测是多个线程同时操作相同的数据，并发插入导致的死锁。
四、并发操作日志
elk上查询方法入参日志。果然在发生异常的时间节点上，发现了两条插入相同数据的日志。相同时间点，两个traceId，可见当时发生了并发调用。
找到当时另一个线程的操作日志，日志无异常，数据入库成功。见下面截图。
五、数据库死锁日志验证推测
当天下午上班后，找运维大哥帮忙，找出死锁日志
执行命令 show ENGINE INNODB status
得到信息如下
=====================================
2023-10-12 13:49:17 0x7fea5ae87700 INNODB MONITOR OUTPUT
=====================================
Per second averages calculated from the last 61 seconds
-----------------
BACKGROUND THREAD
-----------------
srv_master_thread loops: 3123131 srv_active, 0 srv_shutdown, 7715544 srv_idle
srv_master_thread log flush and writes: 10838675
----------
SEMAPHORES
----------
OS WAIT ARRAY INFO: reservation count 3905404
OS WAIT ARRAY INFO: signal count 27617015
RW-shared spins 0, rounds 21456656, OS waits 2772950
RW-excl spins 0, rounds 41132727, OS waits 487320
RW-sx spins 227170, rounds 4363782, OS waits 66152
Spin rounds per wait: 21456656.00 RW-shared, 41132727.00 RW-excl, 19.21 RW-sx
------------------------
LATEST DETECTED DEADLOCK
------------------------
2023-10-12 09:40:50 0x7fecedb1e700
*** (1) TRANSACTION:
TRANSACTION 404898333, ACTIVE 0 sec inserting
mysql tables in use 1, locked 1
LOCK WAIT 2 lock struct(s), heap size 1136, 1 row lock(s), undo log entries 1
MySQL thread id 11375251, OS thread handle 140644580488960, query id 1031478425 10.128.0.160 emax_base update
INSERT INTO emax_user_tag( user_id,
id_card_no,
tag_name,
create_time,
update_time )VALUES( 1712282059297107970,
'A978B69811B82F4B08F929D26B360F2D980CDCB5F419F1DD46527FB90BE1E5A9',
'市场推广',
'2023-10-12 09:40:50.481',
'2023-10-12 09:40:50.481' )
*** (1) WAITING FOR THIS LOCK TO BE GRANTED:
RECORD LOCKS space id 1155 page no 65295 n bits 312 index uniq_user_id_tag_name of table `emax_base`.`emax_user_tag` trx id 404898333 lock mode S waiting
Record lock, heap no 242 PHYSICAL RECORD: n_fields 3; compact format; info bits 0
0: len 8; hex 97c33f75fbadb002; asc ?u ;;
1: len 12; hex e5b882e59cbae68ea8e5b9bf; asc ;;
2: len 8; hex 000000000044c0da; asc D;;
*** (2) TRANSACTION:
TRANSACTION 404898332, ACTIVE 0 sec inserting
mysql tables in use 1, locked 1
3 lock struct(s), heap size 1136, 2 row lock(s), undo log entries 2
MySQL thread id 11375224, OS thread handle 140655576868608, query id 1031478426 10.128.0.160 emax_base update
INSERT INTO emax_user_tag( user_id,
id_card_no,
tag_name,
create_time,
update_time )VALUES( 1712282059297107970,
'A978B69811B82F4B08F929D26B360F2D980CDCB5F419F1DD46527FB90BE1E5A9',
'信息技术服务',
'2023-10-12 09:40:50.481',
'2023-10-12 09:40:50.481' )
*** (2) HOLDS THE LOCK(S):
RECORD LOCKS space id 1155 page no 65295 n bits 312 index uniq_user_id_tag_name of table `emax_base`.`emax_user_tag` trx id 404898332 lock_mode X locks rec but not gap
Record lock, heap no 242 PHYSICAL RECORD: n_fields 3; compact format; info bits 0
0: len 8; hex 97c33f75fbadb002; asc ?u ;;
1: len 12; hex e5b882e59cbae68ea8e5b9bf; asc ;;
2: len 8; hex 000000000044c0da; asc D;;
*** (2) WAITING FOR THIS LOCK TO BE GRANTED:
RECORD LOCKS space id 1155 page no 65295 n bits 312 index uniq_user_id_tag_name of table `emax_base`.`emax_user_tag` trx id 404898332 lock_mode X locks gap before rec insert intention waiting
Record lock, heap no 242 PHYSICAL RECORD: n_fields 3; compact format; info bits 0
0: len 8; hex 97c33f75fbadb002; asc ?u ;;
1: len 12; hex e5b882e59cbae68ea8e5b9bf; asc ;;
2: len 8; hex 000000000044c0da; asc D;;
*** WE ROLL BACK TRANSACTION (1)
<br>此处省略掉若干非关键信息行
----------------------------
END OF INNODB MONITOR OUTPUT
============================
该日志信息量很多，已省略掉若干非关键信息行，我们主要关注最后一次死锁相关信息。
六、死锁复现
sql准备
INSERT INTO emax_user_tag ( user_id, id_card_no, tag_name, create_time, update_time ) VALUES ( 1712282059297107970, 'A978B69811B82F4B08F929D26B360F2D980CDCB5F419F1DD46527FB90BE1E5A9', '市场推广', '2023-10-12 09:40:50.481', '2023-10-12 09:40:50.481' );
INSERT INTO emax_user_tag ( user_id, id_card_no, tag_name, create_time, update_time ) VALUES ( 1712282059297107970, 'A978B69811B82F4B08F929D26B360F2D980CDCB5F419F1DD46527FB90BE1E5A9', '信息技术服务', '2023-10-12 09:40:50.481', '2023-10-12 09:40:50.481' );
本地创建一个和生产一样的数据库emax_user_tag，同样，在user_id和tag_name两个字段上创建唯一索引。
操作步骤1
在窗口一执行
begin;
INSERT INTO emax_user_tag ( user_id, id_card_no, tag_name, create_time, update_time ) VALUES ( 1712282059297107970, 'A978B69811B82F4B08F929D26B360F2D980CDCB5F419F1DD46527FB90BE1E5A9', '市场推广', '2023-10-12 09:40:50.481', '2023-10-12 09:40:50.481' );
操作步骤2
在窗口二执行
begin;
INSERT INTO emax_user_tag ( user_id, id_card_no, tag_name, create_time, update_time ) VALUES ( 1712282059297107970, 'A978B69811B82F4B08F929D26B360F2D980CDCB5F419F1DD46527FB90BE1E5A9', '市场推广', '2023-10-12 09:40:50.481', '2023-10-12 09:40:50.481' );
操作步骤3
在窗口一执行
INSERT INTO emax_user_tag ( user_id, id_card_no, tag_name, create_time, update_time ) VALUES ( 1712282059297107970, 'A978B69811B82F4B08F929D26B360F2D980CDCB5F419F1DD46527FB90BE1E5A9', '信息技术服务', '2023-10-12 09:40:50.481', '2023-10-12 09:40:50.481' );
commit;
操作步骤4
切回窗口二
操作步骤5
执行命令 show ENGINE INNODB status 查看死锁日志。与上面的情况一样。
=====================================
2023-10-13 14:52:43 0x7fe759575700 INNODB MONITOR OUTPUT
=====================================
Per second averages calculated from the last 24 seconds
-----------------
BACKGROUND THREAD
-----------------
srv_master_thread loops: 137590 srv_active, 0 srv_shutdown, 2010475 srv_idle
srv_master_thread log flush and writes: 2148065
----------
SEMAPHORES
----------
OS WAIT ARRAY INFO: reservation count 6627679
OS WAIT ARRAY INFO: signal count 14339066
RW-shared spins 0, rounds 22104735, OS waits 1859432
RW-excl spins 0, rounds 258795010, OS waits 2238669
RW-sx spins 6551822, rounds 136672247, OS waits 1786313
Spin rounds per wait: 22104735.00 RW-shared, 258795010.00 RW-excl, 20.86 RW-sx
------------------------
LATEST DETECTED DEADLOCK
------------------------
2023-10-13 14:50:31 0x7fe892ad0700
*** (1) TRANSACTION:
TRANSACTION 293171186, ACTIVE 10 sec inserting
mysql tables in use 1, locked 1
LOCK WAIT 2 lock struct(s), heap size 1136, 1 row lock(s), undo log entries 1
MySQL thread id 1266048, OS thread handle 140631613069056, query id 35172347 192.168.48.179 develop update
/* ApplicationName=DBeaver 22.2.1 - SQLEditor <Script-4.sql> */ INSERT INTO emax_user_tag ( user_id, id_card_no, tag_name, create_time, update_time ) VALUES ( 1712282059297107970, 'A978B69811B82F4B08F929D26B360F2D980CDCB5F419F1DD46527FB90BE1E5A9', '市场推广', '2023-10-12 09:40:50.481', '2023-10-12 09:40:50.481' )
*** (1) WAITING FOR THIS LOCK TO BE GRANTED:
RECORD LOCKS space id 14171 page no 4 n bits 72 index user_id_tag_name of table `dbsyncer_mf`.`emax_user_tag` trx id 293171186 lock mode S waiting
Record lock, heap no 2 PHYSICAL RECORD: n_fields 3; compact format; info bits 0
0: len 8; hex 97c33f75fbadb002; asc ?u ;;
1: len 12; hex e5b882e59cbae68ea8e5b9bf; asc ;;
2: len 8; hex 80000000000358d0; asc X ;;
*** (2) TRANSACTION:
TRANSACTION 293171182, ACTIVE 19 sec inserting
mysql tables in use 1, locked 1
3 lock struct(s), heap size 1136, 2 row lock(s), undo log entries 2
MySQL thread id 1266036, OS thread handle 140636869953280, query id 35172492 192.168.48.179 develop update
/* ApplicationName=DBeaver 22.2.1 - SQLEditor <Script-1.sql> */ INSERT INTO emax_user_tag ( user_id, id_card_no, tag_name, create_time, update_time ) VALUES ( 1712282059297107970, 'A978B69811B82F4B08F929D26B360F2D980CDCB5F419F1DD46527FB90BE1E5A9', '信息技术服务', '2023-10-12 09:40:50.481', '2023-10-12 09:40:50.481' )
*** (2) HOLDS THE LOCK(S):
RECORD LOCKS space id 14171 page no 4 n bits 72 index user_id_tag_name of table `dbsyncer_mf`.`emax_user_tag` trx id 293171182 lock_mode X locks rec but not gap
Record lock, heap no 2 PHYSICAL RECORD: n_fields 3; compact format; info bits 0
0: len 8; hex 97c33f75fbadb002; asc ?u ;;
1: len 12; hex e5b882e59cbae68ea8e5b9bf; asc ;;
2: len 8; hex 80000000000358d0; asc X ;;
*** (2) WAITING FOR THIS LOCK TO BE GRANTED:
RECORD LOCKS space id 14171 page no 4 n bits 72 index user_id_tag_name of table `dbsyncer_mf`.`emax_user_tag` trx id 293171182 lock_mode X locks gap before rec insert intention waiting
Record lock, heap no 2 PHYSICAL RECORD: n_fields 3; compact format; info bits 0
0: len 8; hex 97c33f75fbadb002; asc ?u ;;
1: len 12; hex e5b882e59cbae68ea8e5b9bf; asc ;;
2: len 8; hex 80000000000358d0; asc X ;;
*** WE ROLL BACK TRANSACTION (1)View Code
七、死锁分析
根据死锁日志进行分析。
执行操作步骤一时：
事务一 获取到了插入意向锁
执行操作步骤二时：
事务二 将事务的意向锁升级为唯一索引排他锁，并且尝试获取唯一索引共享锁（还没有获得共享锁，排队中）
执行操作步骤三时：
事务一 尝试获取插入意向间隙锁
事务二等待事务一释放排他锁
事务一等待事务二释放共享锁
形成相互等待关系，死锁。
八、总结归纳
1.批量插入操作尽量保证数据有序性
2.可借助性能更高的redis进行并发拦截 或 同步处理控制
3.将参数 innodb_deadlock_detect 设置为 on，死锁时会进行回滚（数据库默认开启）
4.业务代码需要考虑数据库异常
5.当前mysql数据库版本是5.7.28-log。经细心的红洁同学验证，mysql-8.0版本针对这个场景做了优化，不会发生死锁。
文献参考 https://blog.csdn.net/minghao0508/article/details/129093202
来源:https://www.cnblogs.com/buguge/p/17788113.html
免责声明：由于采集信息均来自互联网，如果侵犯了您的权益，请联系我们【E-Mail:cb@itdo.tech】 我们会及时删除侵权内容，谢谢合作！
页:
[1]