GreatSQL 死锁案例分析

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

sql 复制代码
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

sql 复制代码
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:

sql 复制代码
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

sql 复制代码
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:

sql 复制代码
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

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

查看锁信息:

sql 复制代码
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

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

查看锁信息:

sql 复制代码
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

sql 复制代码
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:

sql 复制代码
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

sql 复制代码
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:

sql 复制代码
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

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

查看锁信息:

sql 复制代码
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

sql 复制代码
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:

sql 复制代码
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。

Enjoy GreatSQL :)

关于 GreatSQL

GreatSQL是适用于金融级应用的国内自主开源数据库,具备高性能、高可靠、高易用性、高安全等多个核心特性,可以作为MySQL或Percona Server的可选替换,用于线上生产环境,且完全免费并兼容MySQL或Percona Server。

相关链接: GreatSQL社区 Gitee GitHub Bilibili

GreatSQL社区:

社区有奖建议反馈: greatsql.cn/thread-54-1...

社区博客有奖征稿详情: greatsql.cn/thread-100-...

(对文章有疑问或者有独到见解都可以去社区官网提出或分享哦~)

技术交流群:

微信&QQ群:

QQ群:533341697

微信群:添加GreatSQL社区助手(微信号:wanlidbc )好友,待社区助手拉您进群。

相关推荐
苏瞳儿4 小时前
java对数据库的增删改查
java·数据库·oracle
Greyson18 小时前
Layui表格如何使用第三方插件实现树形展示.txt
jvm·数据库·python
2401_871696528 小时前
mysql行级锁失效的原因排查_检查查询条件与执行计划
jvm·数据库·python
Elastic 中国社区官方博客8 小时前
Elasticsearch:快速近似 ES|QL - 第一部分
大数据·运维·数据库·elasticsearch·搜索引擎·全文检索
Dontla8 小时前
高基数(High Cardinality)问题介绍(Prometheus、高基数字段、低基数字段)
前端·数据库·prometheus
a9511416428 小时前
CSS如何实现元素隐藏不占位_使用display-none完全移除
jvm·数据库·python
SelectDB技术团队9 小时前
SelectDB Enterprise 4.0.5:强化安全与治理,构建企业级实时分析与 AI 数据底座
数据库·人工智能·apache doris
一 乐9 小时前
医院挂号|基于springboot + vue医院挂号管理系统(源码+数据库+文档)
java·数据库·vue.js·spring boot·论文·毕设·医院挂号管理系统
ego.iblacat10 小时前
Redis 核心概念与部署
数据库·redis·缓存
m0_4939345310 小时前
如何监控AWR数据收集Job_DBA_SCHEDULER_JOBS中的BSLN_MAINTAIN_STATS
jvm·数据库·python