具有相同查询的死锁

我正试图解决涉及我们繁忙的表之一的死锁的错误。我读过关于死锁的this SO question，虽然它很有意义，但查询顺序似乎不是我的原因。具有相同查询的死锁

这里是SHOW ENGINE INNODB STATUS;的缩略的输出：

*** (1) TRANSACTION: 
TRANSACTION 1 2611184895, ACTIVE 0 sec, process no 17501, OS thread id 140516779579136 starting index read 
mysql tables in use 1, locked 1 
LOCK WAIT 2 lock struct(s), heap size 368, 1 row lock(s) 
MySQL thread id 211935717, query id 3146186174 [SERVER A] Searching rows for update 

UPDATE images_unread_comments 
    SET unread = 0 
    WHERE user_id = 1 AND comment_id IN(1,2,3) AND unread = 1 

*** (1) WAITING FOR THIS LOCK TO BE GRANTED: 
RECORD LOCKS space id 0 page no 404976 n bits 632 index `users_unread_comments` of table images_unread_comments trx id 1 2611184895 lock_mode X waiting 
Record lock, heap no 558 PHYSICAL RECORD: n_fields 3; compact format; info bits 32 
0: len 4; hex 0001461a; asc F ;; 1: len 1; hex 01; asc ;; 2: len 6; hex 00000e67d888; asc g ;; 

*** (2) TRANSACTION: 
TRANSACTION 1 2611184892, ACTIVE 0 sec, process no 17501, OS thread id 140516774520576 updating or deleting, thread declared inside InnoDB 494 
mysql tables in use 1, locked 1 
6 lock struct(s), heap size 1216, 11 row lock(s), undo log entries 1 
MySQL thread id 211935715, query id 3146186169 [SERVER B] Updating 

UPDATE images_unread_comments 
    SET unread = 0 
    WHERE user_id = 1 AND comment_id IN(1,2,3) AND unread = 1 
    *** (2) HOLDS THE LOCK(S): 
RECORD LOCKS space id 0 page no 404976 n bits 632 index users_unread_comments of table images_unread_comments trx id 1 2611184892 lock_mode X 
Record lock, heap no 1 PHYSICAL RECORD: n_fields 1; compact format; info bits 0 
0: len 8; hex 73757072656d756d; asc supremum;; 

Record lock, heap no 555 PHYSICAL RECORD: n_fields 3; compact format; info bits 0 
0: len 4; hex 0001461a; asc F ;; 1: len 1; hex 01; asc ;; 2: len 6; hex 00000e67daf0; asc g ;; 

Record lock, heap no 556 PHYSICAL RECORD: n_fields 3; compact format; info bits 0 
0: len 4; hex 0001461a; asc F ;; 1: len 1; hex 01; asc ;; 2: len 6; hex 00000e67dadb; asc g ;; 

Record lock, heap no 557 PHYSICAL RECORD: n_fields 3; compact format; info bits 0 
0: len 4; hex 0001461a; asc F ;; 1: len 1; hex 01; asc ;; 2: len 6; hex 00000e67d940; asc g @;; 

Record lock, heap no 558 PHYSICAL RECORD: n_fields 3; compact format; info bits 32 
0: len 4; hex 0001461a; asc F ;; 1: len 1; hex 01; asc ;; 2: len 6; hex 00000e67d888; asc g ;; 

*** (2) WAITING FOR THIS LOCK TO BE GRANTED: 
RECORD LOCKS space id 0 page no 404976 n bits 632 index users_unread_comments of table images_unread_comments trx id 1 2611184892 lock_mode X locks gap before rec insert intention waiting 
Record lock, heap no 558 PHYSICAL RECORD: n_fields 3; compact format; info bits 32 
0: len 4; hex 0001461a; asc F ;; 1: len 1; hex 01; asc ;; 2: len 6; hex 00000e67d888; asc g ;; 

*** WE ROLL BACK TRANSACTION (1)

我注意到的事情是两个SQL语句是相同的;但是一个正在服务器A上执行，另一个正在服务器B上执行。不管为什么会发生这种情况 - 如果两个查询都以相同顺序锁定相同的密钥，为什么会造成死锁？还是我首先误解了僵局？

来源

2013-07-04 Graham

在另一个查询（在一个或两个线程中）之前运行的另一个查询很可能已锁定某些行。此外，您还会缩短状态报告的时间 - 第二个等待的事务和它持有的锁定是什么？ – Vatev

@Vatev我已经将该部分的其余部分添加到输出中。 – Graham

我应该注意到这是第一次...锁在另一张桌子上（images_unread_comments）。是否有与这两个表相关的触发器或外键？ – Vatev

看来，事务1已经执行了另一个操作（插入？），其中它锁定了索引中的间隙。它比等待事务2执行更新，因为2已经锁定了ID为1的记录。但事务2无法继续，因为事务1对索引持有锁。如果可以通过此操作隔离事务中使用的所有SQL语句，我们可以看到造成死锁的确切原因

来源

2013-07-06 18:25:43

具有相同查询的死锁

回答

相关问题