使用过Mysql数据库朋友,一定听过读写分离,听的多的,估计耳朵都起茧子了。那么读写分离是怎么实现的呢,最常见的方法就是搭建Mysql的主从复制,主库提供写操作,从库提供读操作,从而达到应用的读写分离。
对于刚入坑开发岗,运维岗的萌新们,一定要弄懂什么是读写分离,读写分离解决什么业务问题,只有彻底弄明白这些之后,才去用读写分离架构。
废话就不多说了,在这里就来说说,主从复制最常见的2种错误
第一种:主键冲突(Error_code: 1062)
第二种:记录丢失,例如update,delete操作,在从库找不到对应记录(Error_code: 1032)
下面来详细模拟一下记录丢失,处理全过程
检查主从复制是否正常
[root@localhost] 11:34:29 [testdb]>show slave status\G;*************************** 1. row *************************** Slave_IO_State: Waiting for master to send event Master_Host: 192.168.0.1 Master_User: repl Master_Port: 3306 Connect_Retry: 60 Master_Log_File: binlog.000029 Read_Master_Log_Pos: 3683 Relay_Log_File: mysql-relay-bin.000003 Relay_Log_Pos: 2207 Relay_Master_Log_File: binlog.000029 Slave_IO_Running: Yes Slave_SQL_Running: Yes
可以看到IO线程和SQL线程运行都是正常的。
创建测试表和记录
[root@localhost] 11:25:48 [testdb]>show create table test1\G;*************************** 1. row *************************** Table: test1Create Table: CREATE TABLE `test1` ( `id` int(11) NOT NULL, `name1` char(10) COLLATE utf8mb4_unicode_ci NOT NULL DEFAULT '', `name2` char(20) COLLATE utf8mb4_unicode_ci NOT NULL DEFAULT '', PRIMARY KEY (`id`)) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci1 row in set (0.07 sec)insert into test1 values(1,'test1','test1');insert into test1 values(2,'test2','test2');insert into test1 values(3,'test3','test3');
模拟主从复制由于从库记录缺失,导致主从复制失败
第一步:在从库中删除id=2的记录
[root@localhost] 11:26:41 [testdb]>delete from test1 where id=2;Query OK, 1 row affected (0.44 sec)[root@localhost] 11:26:52 [testdb]>select * from test1;+----+-------+-------+| id | name1 | name2 |+----+-------+-------+| 1 | test1 | test1 || 3 | test3 | test3 |+----+-------+-------+2 rows in set (0.00 sec)
第二步:在主库上删除id=2的记录
[root@localhost] 11:27:11 [testdb]>delete from test1 where id=2;Query OK, 1 row affected (0.17 sec)[root@localhost] 11:27:51 [testdb]>select * from test1;+----+-------+-------+| id | name1 | name2 |+----+-------+-------+| 1 | test1 | test1 || 3 | test3 | test3 |+----+-------+-------+2 rows in set (0.00 sec)
在从库上查看主从复制情况
[root@localhost] 11:34:05 [testdb]>show slave status\G;*************************** 1. row *************************** Slave_IO_State: Waiting for master to send event Master_Host: 192.168.0.1 Master_User: repl Master_Port: 3306 Connect_Retry: 60 Master_Log_File: binlog.000029 Read_Master_Log_Pos: 3683 Relay_Log_File: mysql-relay-bin.000003 Relay_Log_Pos: 1929 Relay_Master_Log_File: binlog.000029 Slave_IO_Running: Yes Slave_SQL_Running: No Replicate_Do_DB: Replicate_Ignore_DB: Replicate_Do_Table: Replicate_Ignore_Table: Replicate_Wild_Do_Table: Replicate_Wild_Ignore_Table: Last_Errno: 1032 Last_Error: Could not execute Delete_rows event on table testdb.test1; Can't find record in 'test1', Error_code: 1032; handler error HA_ERR_KEY_NOT_FOUND; the event's master log binlog.000029, end_log_pos 3652 Skip_Counter: 0 Exec_Master_Log_Pos: 3405 Relay_Log_Space: 2414 Until_Condition: None Until_Log_File: Until_Log_Pos: 0 Master_SSL_Allowed: No Master_SSL_CA_File: Master_SSL_CA_Path: Master_SSL_Cert: Master_SSL_Cipher: Master_SSL_Key: Seconds_Behind_Master: NULLMaster_SSL_Verify_Server_Cert: No Last_IO_Errno: 0 Last_IO_Error: Last_SQL_Errno: 1032 Last_SQL_Error: Could not execute Delete_rows event on table testdb.test1; Can't find record in 'test1', Error_code: 1032; handler error HA_ERR_KEY_NOT_FOUND; the event's master log binlog.000029, end_log_pos 3652 Replicate_Ignore_Server_Ids: Master_Server_Id: 111213106 Master_UUID: 3ada166e-c4db-11ea-b21d-000c29cc2388 Master_Info_File: mysql.slave_master_info SQL_Delay: 0 SQL_Remaining_Delay: NULL Slave_SQL_Running_State: Master_Retry_Count: 86400 Master_Bind: Last_IO_Error_Timestamp: Last_SQL_Error_Timestamp: 200904 11:33:10 Master_SSL_Crl: Master_SSL_Crlpath: Retrieved_Gtid_Set: 3ada166e-c4db-11ea-b21d-000c29cc2388:84830-84835 Executed_Gtid_Set: 3ada166e-c4db-11ea-b21d-000c29cc2388:1-84834,3ada166e-c4db-11ea-b21d-000c29cc2389:1-4 Auto_Position: 1 Replicate_Rewrite_DB: Channel_Name: Master_TLS_Version:1 row in set (0.00 sec)
此时主从的sql线程已经是停止状态,主从复制的数据已经不同步了。复制开始报1032错误了。
要解决1032错误,可以有以下3中方案
方案一:手工将缺失的业务记录在主库上导出,并导入到从库,然后启动从库的sql线程就可以了。慢着,大家有没有注意到一个问题,就是在主库上,到底要导出哪条记录,报错信息里并没有,但是有提示,he event's master log binlog.000029, end_log_pos 3652,所以还需要将binlog日志里的内容解析处理,找到要操作的记录,似乎有些麻烦。不用慌,还有方案二,方案三。
方案二:Mysql数据库提供一个参数slave_skip_errors,这个参数可以跳过指定错误代码的sql语句,例如:slave_skip_errors=1032,可惜,这个参数不能在线修改,修改生效需要重启实例,是不是也太友好。
[root@localhost] 11:28:57 [testdb]>set global slave_skip_errors=1032;ERROR 1238 (HY000): Variable 'slave_skip_errors' is a read only variable
方案三:使用percona-toolkits工具集中的pt-slave-restart工具,自动跳过主从同步指定的报错代码sql语句,此方法对mysql数据侵入性小,不必重启Mysql实例
[mysql@mysql ~]$ pt-slave-restart --user=root --password=root --socket=/data/mysql/run/3306/mysql.sock --error-numbers=1032# A software update is available:2020-09-04T11:32:07 S=/data/mysql/run/3306/mysql.sock,p=...,u=root mysql-relay-bin.000003 1651 1032
当跳过主从同步指定的报错代码sql语句,主从复制恢复之后,间隔64秒,会再次自动检测主从复制是否有1032错误。
其它类似的错误,都可以用以上三种方法方案处理,建议大家使用方案三。