千家信息网

【Mysql】两条insert 语句产生的死锁

发表于:2024-10-25 作者:千家信息网编辑
千家信息网最后更新 2024年10月25日,背景:查看status日志发现两条insert 出现了死锁RECORD LOCKS space id 388 page no 27032 n bits 616 index `idx_svcorders
千家信息网最后更新 2024年10月25日【Mysql】两条insert 语句产生的死锁背景:查看status日志发现两条insert 出现了死锁
  1. RECORD LOCKS space id 388 page no 27032 n bits 616 index `idx_svcorderserviceitem_workorderid_quantity` of table `ecejservice`.`svc_order_service_item` trx id 596252578 lock_mode X insert intention waiting
可以确定,这个x锁不是由于INSERT产生的,因为 INSERT可能产生的锁包括检查dup key时的s锁,隐式锁转换为显式锁(not gap,要在二级索引上产生lock_mode为X的LOCK_ORDINARY类型的锁(包括记录及记录前面的gap),据我所知一般是根据二级索引扫描进行记录更新导致的。

3. 根据
LOCK WAIT 14 lock struct(s), heap size 2936, 7 row lock(s), undo log entries 7
有7个undo entires,而单纯的INSERT一条记录只有一个undo entry,因此可以推断除了INSERT,必然还有别的操作

基于以上,事务除了INSERT,可能还存在DELETE/UPDATE,并且这些操作是走的二级索引来查找更新记录。

一个简单但不完全相同的重现步骤: DROP TABLE t1; CREATE TABLE `t1` ( `a` int(11) NOT NULL AUTO_INCREMENT, `b` int(11) DEFAULT NULL, `c` int(11) DEFAULT NULL, PRIMARY KEY (`a`), KEY `b` (`b`) ) ENGINE=InnoDB ; insert into t1(a, b,c) values(1,2,3),(5,4,6),(8, 7,9),(12,12,19),(15,15,11);

session1:

begin; delete from t1 where b = 12; //二级索引上lock_mode X、lock_mode X locks gap before rec以及主键上的lock_mode X locks rec but not gap 二级索引:heap_no=5, type_mode=3 (12上的LOCK_ORDINARY类型锁,包括记录和记录前的GAP) 聚集索引:heap_no=5,type_mode=1027 二级索引:heap_no=6,type_mode=547(15上的GAP锁) session2: begin; delete from t1 where b = 7; //二级索引上lock_mode X、lock_mode X locks gap before rec以及主键上的lock_mode X locks rec but not gap 二级索引:heap_no=4,type_mode=3 (7上的LOCK_ORDINARY类型锁,包括记录和记录前的GAP) 聚集索引:heap_no=4,type_mode=1027 二级索引:heap_no=5,type_mode=547 (记录12上的GAP锁) session1: insert into t1 values (NULL, 6,10); //新插入记录聚集索引无冲突插入成功,二级索引等待插入意向锁(lock_mode X locks gap before rec insert intention waiting) 二级索引,heap_no=4, type_mode=2819 (请求记录7上面的插入意向锁LOCK_X | LOCK_GAP | LOCK_INSERT_INTENTION, 需要等待session2 session2: insert into t1 values (NULL, 7,10); 二级索引:heap_no=5, type_mode=2819 (请求记录12上的插入意向锁LOCK_X | LOCK_GAP | LOCK_INSERT_INTENTION,需要等待session1) 互相等待,导致发生死锁 从打印的死锁信息来看,基本和线上发生的死锁现象是一致的。


再举一个例子
  1. mysql> select * from test01;
  2. +----+-----+
  3. | id | app |
  4. +----+-----+
  5. | 1 | 01 |
  6. | 2 | 02 |
  7. | 5 | 03 |
  8. | 10 | 03 |
  9. | 6 | 04 |
  10. | 7 | 05 |
  11. | 8 | 06 |
  12. | 9 | 06 |
  13. | 11 | 06 |
  14. | 12 | 07 |
  15. | 13 | 08 |
  16. | 14 | 09 |
  17. | 15 | 09 |
  18. +----+-----+
  19. 13 rows in set (0.00 sec)

  • session1:
  • mysql> select now();start TRANSACTION;
  • +---------------------+
  • | now() |
  • +---------------------+
  • | 2018-01-25 16:08:46 |
  • +---------------------+
  • 1 row in set (0.00 sec)

  • Query OK, 0 rows affected (0.00 sec)

  • mysql> select * from test01 where app='05' for update; ---第1步 锁住【6.04】-【7.05】以及【7.05】-【8.06】 两段区间
  • +----+-----+
  • | id | app |
  • +----+-----+
  • | 7 | 05 |
  • +----+-----+
  • 1 row in set (0.00 sec)

  • mysql> insert into test01(app) values ('07'); --第三步 等待第二步释放
  • Query OK, 1 row affected (23.24 sec)



  • session2:
  • mysql> select * from test01 where app='08' for update; --第二步 锁住【12,07】-【13,08】以及【13,08】-【14,09】两段区间
  • +----+-----+
  • | id | app |
  • +----+-----+
  • | 13 | 08 |
  • +----+-----+
  • 1 row in set (0.00 sec)

  • mysql> insert into test01(app) values ('04'); ----第四步 等待第一步释放,,于是死锁
  • ERROR 1213 (40001): Deadlock found when trying to get lock; try restarting transaction


  • 查看死锁日志:
    1. LATEST DETECTED DEADLOCK
    2. ------------------------
    3. 2018-01-25 16:09:54 0x7f07d23ff700
    4. *** (1) TRANSACTION:
    5. TRANSACTION 5375, ACTIVE 51 sec inserting
    6. mysql tables in use 1, locked 1
    7. LOCK WAIT 5 lock struct(s), heap size 1136, 4 row lock(s), undo log entries 1
    8. MySQL thread id 2294, OS thread handle 139671567841024, query id 42463 localhost root update
    9. insert into test01(app) values ('07')
    10. *** (1) WAITING FOR THIS LOCK TO BE GRANTED:
    11. RECORD LOCKS space id 64 page no 4 n bits 80 index idx_app of table `devops`.`test01` trx id 5375 lock_mode X locks gap before rec insert intention waiting
    12. Record lock, heap no 12 PHYSICAL RECORD: n_fields 2; compact format; info bits 0
    13. 0: len 2; hex 3038; asc 08;;
    14. 1: len 4; hex 0000000d; asc ;;

    15. *** (2) TRANSACTION:
    16. TRANSACTION 5376, ACTIVE 38 sec inserting
    17. mysql tables in use 1, locked 1
    18. 5 lock struct(s), heap size 1136, 4 row lock(s), undo log entries 1
    19. MySQL thread id 2293, OS thread handle 139671568905984, query id 42464 localhost root update
    20. insert into test01(app) values ('04')
    1. *** (2) HOLDS THE LOCK(S):
    1. RECORD LOCKS space id 64 page no 4 n bits 80 index idx_app of table `devops`.`test01` trx id 5376 lock_mode X
    2. Record lock, heap no 12 PHYSICAL RECORD: n_fields 2; compact format; info bits 0
    3. 0: len 2; hex 3038; asc 08;;
    4. 1: len 4; hex 0000000d; asc ;;

    5. *** (2) WAITING FOR THIS LOCK TO BE GRANTED:
    6. RECORD LOCKS space id 64 page no 4 n bits 80 index idx_app of table `devops`.`test01` trx id 5376 lock_mode X locks gap before rec insert intention waiting
    7. Record lock, heap no 6 PHYSICAL RECORD: n_fields 2; compact format; info bits 0
    8. 0: len 2; hex 3035; asc 05;;
    9. 1: len 4; hex 00000007; asc ;;

    10. *** WE ROLL BACK TRANSACTION (2)
    11. ------------

    死锁日志是不是和上面的一样?



    参考:
    http://blog.itpub.net/22664653/viewspace-2145068/ ----杨奇龙
    http://www.sohu.com/a/169663059_610509 ---insert ..select 语句产生死锁
    http://blog.itpub.net/7728585/viewspace-2146183/ ---insert ..select 语句产生死锁--八怪
    0