mysql鎖之死鎖

    死鎖是事務繞不開的話題，mysql當然也不例外，本文主要模擬一下mysql的死鎖，以及應對措施。
    首先看一個引數，預設innodb_print_all_deadlocks引數是關閉。開啟後可以將死鎖記錄到error.log中。否則只能通過show engine innodb status檢視。
mysql> SHOW VARIABLES LIKE 'INNODB_PRINT_ALL_DEADLOCKS';
+----------------------------+-------+
| Variable_name              | Value |
+----------------------------+-------+
| innodb_print_all_deadlocks | OFF   |
+----------------------------+-------+
1 row in set (0.00 sec)

開啟innodb_print_all_deadlocks,改引數是全域性引數，可以動態調整。
mysql> SET GLOBAL innodb_print_all_deadlocks=1;
Query OK, 0 rows affected (0.00 sec)

表test01上c1是主鍵，c2是唯一約束。
mysql> show create table test01\G
*************************** 1. row ***************************
       Table: test01
Create Table: CREATE TABLE `test01` (
  `c1` bigint(20) NOT NULL AUTO_INCREMENT,
  `c2` int(11) DEFAULT NULL,
  PRIMARY KEY (`c1`),
  UNIQUE KEY `uidx_test01_c2` (`c2`)
) ENGINE=InnoDB AUTO_INCREMENT=4 DEFAULT CHARSET=utf8
1 row in set (0.00 sec)
mysql> select * from test01;
+----+------+
| c1 | c2   |
+----+------+
|  1 |    1 |
|  2 |    2 |
|  3 |    3 |
+----+------+
3 rows in set (0.00 sec)
會話A
mysql> begin ;
Query OK, 0 rows affected (0.00 sec)

mysql> delete from test01 where c2=2;
Query OK, 1 row affected (0.00 sec)

會話B
mysql> begin;
Query OK, 0 rows affected (0.00 sec)

mysql> delete from test01 where c2=2;
ERROR 1213 (40001): Deadlock found when trying to get lock; try restarting transaction

會話A
mysql> insert into test01 select 2,2;
Query OK, 1 row affected (0.00 sec)
Records: 1  Duplicates: 0  Warnings: 0
會話A在（2,2）上加了X的行鎖，會話B要刪除相同的資料行，那麼也要在該行上加X的行鎖（lock_mode X locks rec but not gap），所以出現了等待（lock_mode X waiting）。後面會話A要插入一行（2,2），因為欄位c2上有唯一索引，插入的時候要檢查duplicate key的檢查，這個過程需要申請S的鎖，而在得到這個鎖之前，它需要等會話B先得到會話A最開始執行的X鎖。也就是說，會話B要等待會話A第一條語句釋放X鎖，會話A第二條語句又要等待會話B釋放X鎖，兩個會話之間形成了等待的閉合迴路，形成了死鎖。出現死鎖後，mysql會選擇一個小事務進行回滾，以解決死鎖。



show engine innodb status檢視死鎖資訊，error.log中記錄的死鎖也類似下面：
------------------------
LATEST DETECTED DEADLOCK
------------------------
2019-07-18 11:11:32 0x7fdc50298700
*** (1) TRANSACTION:
TRANSACTION 713521, ACTIVE 122 sec starting index read
mysql tables in use 1, locked 1
LOCK WAIT 2 lock struct(s), heap size 1136, 2 row lock(s)
MySQL thread id 4, OS thread handle 140584214165248, query id 144 localhost root updating
delete from test01 where c2=2
*** (1) WAITING FOR THIS LOCK TO BE GRANTED:
RECORD LOCKS space id 25 page no 4 n bits 72 index uidx_test01_c2 of table `ming`.`test01` trx id 713521 lock_mode X waiting
Record lock, heap no 3 PHYSICAL RECORD: n_fields 2; compact format; info bits 32
 0: len 4; hex 80000002; asc     ;;
 1: len 8; hex 8000000000000002; asc         ;;

*** (2) TRANSACTION:
TRANSACTION 713523, ACTIVE 16 sec inserting
mysql tables in use 1, locked 1
4 lock struct(s), heap size 1136, 3 row lock(s), undo log entries 2
MySQL thread id 3, OS thread handle 140584214431488, query id 146 localhost root executing
insert into test01 select 2,2
*** (2) HOLDS THE LOCK(S):
RECORD LOCKS space id 25 page no 4 n bits 72 index uidx_test01_c2 of table `ming`.`test01` trx id 713523 lock_mode X locks rec but not gap
Record lock, heap no 3 PHYSICAL RECORD: n_fields 2; compact format; info bits 32
 0: len 4; hex 80000002; asc     ;;
 1: len 8; hex 8000000000000002; asc         ;;

*** (2) WAITING FOR THIS LOCK TO BE GRANTED:
RECORD LOCKS space id 25 page no 4 n bits 72 index uidx_test01_c2 of table `ming`.`test01` trx id 713523 lock mode S waiting
Record lock, heap no 3 PHYSICAL RECORD: n_fields 2; compact format; info bits 32
 0: len 4; hex 80000002; asc     ;;
 1: len 8; hex 8000000000000002; asc         ;;

針對死鎖的舉措：
1.如果死鎖較多的話，那麼建議開啟innodb_print_all_deadlocks,因為show engine innodb status只顯示最近一次的死鎖資訊。
2.保持事務小而短，並儘快提價
3.避免在一個事務裡面修改幾張表，或者是同一張表修改不同的結果集
4.可以嘗試使用較低的隔離級別，比如RC。也可以使用鎖定讀
5.建立合適的索引
6.如果應用允許的話，序列化事務

來自 “ ITPUB部落格 ” ，連結：http://blog.itpub.net/31480688/viewspace-2651193/，如需轉載，請註明出處，否則將追究法律責任。