GreatSQL 死锁案例分析

2024年 4月 19日 29.6k 0

1.背景概述

客户业务发生死锁的报错,根据业务程序日志及业务流程,发现造成死锁的原因是:事务1 delete + insert ,事务2 delete + insert  2个事务交替执行导致的死锁;由于GAP锁阻塞了插入意向锁,并且当delete的数据存在时死锁不会发生,当delete的数据不存在时,会发生死锁。

2.问题复现

本次测试基于 GreatSQL-8.0.32-24,隔离级别为 RR

2.1 创建测试表

greatsql> create database test;

greatsql> create table test(c1 int unique key, c2 int, c3 int, c4 int);

greatsql> insert into test values (1,1,1,1),(3,3,3,3),(5,5,5,5),(9,9,9,9);

greatsql> select * from test;
+------+------+------+------+
| c1 | c2 | c3 | c4 |
+------+------+------+------+
| 1 | 1 | 1 | 1 |
| 3 | 3 | 3 | 3 |
| 5 | 5 | 5 | 5 |
| 9 | 9 | 9 | 9 |
+------+------+------+------+
4 rows in set (0.01 sec)

2.2 事务执行顺序

按以下事务执行顺序,如果要删除的数据存在,则不会发生死锁;如果要删除的数据不存,并且要删除的数据在同一个GAP锁的区间内则会发生死锁;

时间 事务1 事务2
T1 BEGIN; BEGIN;
T2 delete from test where c1=?;
T3 delete from test where c1=?;
T4 insert into test value(?,?,?,?);
T5 insert into test value(?,?,?,?);

2.3 当delete的数据存在时

事务1:delete

greatsql> begin;
Query OK, 0 rows affected (0.00 sec)

greatsql> delete from test where c1=3;
Query OK, 1 row affected (0.00 sec)

查看锁信息:

greatsql> select THREAD_ID,OBJECT_SCHEMA,OBJECT_NAME,INDEX_NAME,LOCK_TYPE,LOCK_MODE,LOCK_STATUS,LOCK_DATA from performance_schema.data_locks;

+-----------+---------------+-------------+-----------------+-----------+---------------+-------------+-------------------+
| THREAD_ID | OBJECT_SCHEMA | OBJECT_NAME | INDEX_NAME | LOCK_TYPE | LOCK_MODE | LOCK_STATUS | LOCK_DATA |
+-----------+---------------+-------------+-----------------+-----------+---------------+-------------+-------------------+
| 59 | test | test | NULL | TABLE | IX | GRANTED | NULL |
| 59 | test | test | c1 | RECORD | X,REC_NOT_GAP | GRANTED | 3, 0x000000000201 |
| 59 | test | test | GEN_CLUST_INDEX | RECORD | X,REC_NOT_GAP | GRANTED | 0x000000000201 |
+-----------+---------------+-------------+-----------------+-----------+---------------+-------------+-------------------+
3 rows in set (0.00 sec)

此时事务1给 3, 0x000000000201 这条数据加了 记录锁 X,REC_NOT_GAP

事务2:delete

greatsql> begin;
Query OK, 0 rows affected (0.00 sec)

greatsql> delete from test where c1=5;
Query OK, 1 row affected (0.00 sec)

查看锁信息:

greatsql> select THREAD_ID,OBJECT_SCHEMA,OBJECT_NAME,INDEX_NAME,LOCK_TYPE,LOCK_MODE,LOCK_STATUS,LOCK_DATA from performance_schema.data_locks;

+-----------+---------------+-------------+-----------------+-----------+---------------+-------------+-------------------+
| THREAD_ID | OBJECT_SCHEMA | OBJECT_NAME | INDEX_NAME | LOCK_TYPE | LOCK_MODE | LOCK_STATUS | LOCK_DATA |
+-----------+---------------+-------------+-----------------+-----------+---------------+-------------+-------------------+
| 57 | test | test | NULL | TABLE | IX | GRANTED | NULL |
| 57 | test | test | c1 | RECORD | X,REC_NOT_GAP | GRANTED | 5, 0x000000000202 |
| 57 | test | test | GEN_CLUST_INDEX | RECORD | X,REC_NOT_GAP | GRANTED | 0x000000000202 |
| 59 | test | test | NULL | TABLE | IX | GRANTED | NULL |
| 59 | test | test | c1 | RECORD | X,REC_NOT_GAP | GRANTED | 3, 0x000000000201 |
| 59 | test | test | GEN_CLUST_INDEX | RECORD | X,REC_NOT_GAP | GRANTED | 0x000000000201 |
+-----------+---------------+-------------+-----------------+-----------+---------------+-------------+-------------------+
6 rows in set (0.00 sec)

此时事务2给 5, 0x000000000202 这条数据加了 记录锁 X,REC_NOT_GAP

事务1:insert

greatsql> insert into test value(3,3,3,3);
Query OK, 1 row affected (0.00 sec)

查看锁信息:

greatsql> select THREAD_ID,OBJECT_SCHEMA,OBJECT_NAME,INDEX_NAME,LOCK_TYPE,LOCK_MODE,LOCK_STATUS,LOCK_DATA from performance_schema.data_locks;

+-----------+---------------+-------------+-----------------+-----------+---------------+-------------+-------------------+

| THREAD_ID | OBJECT_SCHEMA | OBJECT_NAME | INDEX_NAME | LOCK_TYPE | LOCK_MODE | LOCK_STATUS | LOCK_DATA |

+-----------+---------------+-------------+-----------------+-----------+---------------+-------------+-------------------+

| 57 | test | test | NULL | TABLE | IX | GRANTED | NULL |

| 57 | test | test | c1 | RECORD | X,REC_NOT_GAP | GRANTED | 5, 0x000000000202 |

| 57 | test | test | GEN_CLUST_INDEX | RECORD | X,REC_NOT_GAP | GRANTED | 0x000000000202 |

| 59 | test | test | NULL | TABLE | IX | GRANTED | NULL |

| 59 | test | test | c1 | RECORD | X,REC_NOT_GAP | GRANTED | 3, 0x000000000201 |

| 59 | test | test | GEN_CLUST_INDEX | RECORD | X,REC_NOT_GAP | GRANTED | 0x000000000201 |

| 59 | test | test | c1 | RECORD | S,GAP | GRANTED | 3, 0x000000000201 |

| 59 | test | test | c1 | RECORD | S,GAP | GRANTED | 5, 0x000000000202 |

| 59 | test | test | c1 | RECORD | S,GAP | GRANTED | 3, 0x000000000206 |

+-----------+---------------+-------------+-----------------+-----------+---------------+-------------+-------------------+
9 rows in set (0.00 sec)

此时事务1给 被delete删除的数据 3, 0x000000000201 ,插入的数据 3, 0x000000000202,以及相邻的下一条数据 5, 0x000000000202 加了 间隙锁 S,GAP

事务2:insert

greatsql> insert into test value(5,5,5,5);
Query OK, 1 row affected (0.01 sec)

查看锁信息:

greatsql> select THREAD_ID,OBJECT_SCHEMA,OBJECT_NAME,INDEX_NAME,LOCK_TYPE,LOCK_MODE,LOCK_STATUS,LOCK_DATA from performance_schema.data_locks;

+-----------+---------------+-------------+-----------------+-----------+---------------+-------------+-------------------+
| THREAD_ID | OBJECT_SCHEMA | OBJECT_NAME | INDEX_NAME | LOCK_TYPE | LOCK_MODE | LOCK_STATUS | LOCK_DATA |
+-----------+---------------+-------------+-----------------+-----------+---------------+-------------+-------------------+
| 57 | test | test | NULL | TABLE | IX | GRANTED | NULL |
| 57 | test | test | c1 | RECORD | X,REC_NOT_GAP | GRANTED | 5, 0x000000000202 |
| 57 | test | test | GEN_CLUST_INDEX | RECORD | X,REC_NOT_GAP | GRANTED | 0x000000000202 |
| 57 | test | test | c1 | RECORD | S,GAP | GRANTED | 5, 0x000000000202 |
| 57 | test | test | c1 | RECORD | S,GAP | GRANTED | 9, 0x000000000203 |
| 57 | test | test | c1 | RECORD | S,GAP | GRANTED | 5, 0x000000000207 |
| 59 | test | test | NULL | TABLE | IX | GRANTED | NULL |
| 59 | test | test | c1 | RECORD | X,REC_NOT_GAP | GRANTED | 3, 0x000000000201 |
| 59 | test | test | GEN_CLUST_INDEX | RECORD | X,REC_NOT_GAP | GRANTED | 0x000000000201 |
| 59 | test | test | c1 | RECORD | S,GAP | GRANTED | 3, 0x000000000201 |
| 59 | test | test | c1 | RECORD | S,GAP | GRANTED | 5, 0x000000000202 |
| 59 | test | test | c1 | RECORD | S,GAP | GRANTED | 3, 0x000000000206 |
+-----------+---------------+-------------+-----------------+-----------+---------------+-------------+-------------------+
12 rows in set (0.00 sec)

此时事务2给 被delete删除的数据 5, 0x000000000202 ,插入的数据 5, 0x000000000207,以及相邻的下一条数据 9, 0x000000000203 加了 间隙锁 S,GAP

由于GAP锁之间是相互兼容的,所以没有发生锁等待及死锁,此时事务1,事务2都执行完成,可以正常提交。

2.4 当delete的数据不存在时

事务1:delete

greatsql> begin;
Query OK, 0 rows affected (0.01 sec)

greatsql> delete from test where c1=6;
Query OK, 0 rows affected (0.00 sec)

查看锁信息:

greatsql> select THREAD_ID,OBJECT_SCHEMA,OBJECT_NAME,INDEX_NAME,LOCK_TYPE,LOCK_MODE,LOCK_STATUS,LOCK_DATA from performance_schema.data_locks;

+-----------+---------------+-------------+------------+-----------+-----------+-------------+-------------------+
| THREAD_ID | OBJECT_SCHEMA | OBJECT_NAME | INDEX_NAME | LOCK_TYPE | LOCK_MODE | LOCK_STATUS | LOCK_DATA |
+-----------+---------------+-------------+------------+-----------+-----------+-------------+-------------------+
| 61 | test | test | NULL | TABLE | IX | GRANTED | NULL |
| 61 | test | test | c1 | RECORD | X,GAP | GRANTED | 9, 0x000000000203 |
+-----------+---------------+-------------+------------+-----------+-----------+-------------+-------------------+
2 rows in set (0.00 sec)

此时事务1给 9, 0x000000000203 这条数据加了 间隙锁 X,GAP

事务2:delete

greatsql> begin;
Query OK, 0 rows affected (0.00 sec)

greatsql> delete from test where c1=7;
Query OK, 0 rows affected (0.00 sec)

查看锁信息:

greatsql> select THREAD_ID,OBJECT_SCHEMA,OBJECT_NAME,INDEX_NAME,LOCK_TYPE,LOCK_MODE,LOCK_STATUS,LOCK_DATA from performance_schema.data_locks;
+-----------+---------------+-------------+------------+-----------+-----------+-------------+-------------------+
| THREAD_ID | OBJECT_SCHEMA | OBJECT_NAME | INDEX_NAME | LOCK_TYPE | LOCK_MODE | LOCK_STATUS | LOCK_DATA |
+-----------+---------------+-------------+------------+-----------+-----------+-------------+-------------------+
| 62 | test | test | NULL | TABLE | IX | GRANTED | NULL |
| 62 | test | test | c1 | RECORD | X,GAP | GRANTED | 9, 0x000000000203 |
| 61 | test | test | NULL | TABLE | IX | GRANTED | NULL |
| 61 | test | test | c1 | RECORD | X,GAP | GRANTED | 9, 0x000000000203 |
+-----------+---------------+-------------+------------+-----------+-----------+-------------+-------------------+
4 rows in set (0.00 sec)

此时事务2给 9, 0x000000000203 这条数据加了 间隙锁 X,GAP 间隙锁可以相互兼容,因此没有报错

事务1:insert

greatsql> insert into test value(6,6,6,6);
---hang住,处于锁等待

查看锁信息:

greatsql> select THREAD_ID,OBJECT_SCHEMA,OBJECT_NAME,INDEX_NAME,LOCK_TYPE,LOCK_MODE,LOCK_STATUS,LOCK_DATA from performance_schema.data_locks;

+-----------+---------------+-------------+------------+-----------+------------------------+-------------+-------------------+
| THREAD_ID | OBJECT_SCHEMA | OBJECT_NAME | INDEX_NAME | LOCK_TYPE | LOCK_MODE | LOCK_STATUS | LOCK_DATA |
+-----------+---------------+-------------+------------+-----------+------------------------+-------------+-------------------+
| 62 | test | test | NULL | TABLE | IX | GRANTED | NULL |
| 62 | test | test | c1 | RECORD | X,GAP | GRANTED | 9, 0x000000000203 |
| 61 | test | test | NULL | TABLE | IX | GRANTED | NULL |
| 61 | test | test | c1 | RECORD | X,GAP | GRANTED | 9, 0x000000000203 |
| 61 | test | test | c1 | RECORD | X,GAP,INSERT_INTENTION | WAITING | 9, 0x000000000203 |
+-----------+---------------+-------------+------------+-----------+------------------------+-------------+-------------------+
5 rows in set (0.00 sec)

此时事务1,要申请给 9, 0x000000000203 这条数据加 间隙锁,插入意向锁 X,GAP,INSERT_INTENTION;由于事务2已经加了 间隙锁 X,GAP 间隙锁与插入意向锁并不兼容,因此事务1的insert处于锁等待状态

事务2:insert

greatsql> insert into test value(7,7,7,7);
ERROR 1213 (40001): Deadlock found when trying to get lock; try restarting transaction

查看锁信息:

greatsql> select THREAD_ID,OBJECT_SCHEMA,OBJECT_NAME,INDEX_NAME,LOCK_TYPE,LOCK_MODE,LOCK_STATUS,LOCK_DATA from performance_schema.data_locks;

+-----------+---------------+-------------+-----------------+-----------+------------------------+-------------+------------------------+
| THREAD_ID | OBJECT_SCHEMA | OBJECT_NAME | INDEX_NAME | LOCK_TYPE | LOCK_MODE | LOCK_STATUS | LOCK_DATA |
+-----------+---------------+-------------+-----------------+-----------+------------------------+-------------+------------------------+
| 61 | test | test | NULL | TABLE | IX | GRANTED | NULL |
| 61 | test | test | c1 | RECORD | X,GAP | GRANTED | 9, 0x000000000203 |
| 61 | test | test | c1 | RECORD | X,GAP | GRANTED | 6, 0x000000000213 |
| 61 | test | test | GEN_CLUST_INDEX | RECORD | X | GRANTED | supremum pseudo-record |
| 61 | test | test | GEN_CLUST_INDEX | RECORD | X,GAP | GRANTED | 0x000000000213 |
| 61 | test | test | c1 | RECORD | X,GAP,INSERT_INTENTION | GRANTED | 9, 0x000000000203 |
+-----------+---------------+-------------+-----------------+-----------+------------------------+-------------+------------------------+
6 rows in set (0.01 sec)

事务2回滚,只有事务1的加锁信息。

由于此时事务2,要申请给 9, 0x000000000203 这条数据加 间隙锁,插入意向锁由于事务1已经加了 间隙锁 X,GAP 间隙锁与插入意向锁并不兼容,因此事务2的insert处于锁等待状态,2个事务相互等待锁导致死锁,此时事务2回滚。

3.总结

此次死锁的发生主要是GAP 锁 和 插入意向锁的冲突,建议让业务修改一下逻辑,先判断数据是否存在 select 一下,存在的话 delete 后在 insert ;  不存在的话直接 insert 不用delete了。

相关文章

pt-kill工具的使用
pt-ioprofile工具包的使用
数据库管理-第216期 Oracle的高可用-01(20240703)
DBMS_REPAIR EXAMPLE SCRIPT WITH PARTITION
数据库事务的四大特性: ACID 
使用BBED修复损坏的SYSTEM文件头

发布评论