GreatSQL一个关于主从复制的限制描述与规避

2023-09-27 09:42 由 GreatSQL 发表于 #数据库

一、背景

分享一个在项目运维中遇到的一个主从复制限制的一个坑，项目的架构为主集群+灾备集群，每个集群为一主两从模式。主集群到灾备集群的同步为主从复制的方式，根据业务需求灾备集群需要忽略系统库跟某些配置表，所以才会触发此限制，而这个限制如果我们之前没有遇到过，那么排查起来也是相对不易的。

二、限制描述

1、主从同步出现报错

greatsql> show slave status\G
*************************** 1. row ***************************
               Slave_IO_State: Waiting for master to send event
                  Master_Host: 192.168.xxx.xxx
                  Master_User: repl
                  Master_Port: 3306
                Connect_Retry: 60
              Master_Log_File: greatsql-bin.000990
          Read_Master_Log_Pos: 92274290
               Relay_Log_File: greatsql-relay.002963     -----
                Relay_Log_Pos: 701548899
        Relay_Master_Log_File: greatsql-bin.000988
             Slave_IO_Running: Yes
            Slave_SQL_Running: No
              Replicate_Do_DB:
          Replicate_Ignore_DB: mysql,dbscale,dbscale_tmp,information_schema,performance_schema,sys
           Replicate_Do_Table:
       Replicate_Ignore_Table:
      Replicate_Wild_Do_Table:
  Replicate_Wild_Ignore_Table: A.ab,B.bc
                   Last_Errno: 1146
                   Last_Error: Coordinator stopped because there were error(s) in the worker(s). The most recent failure being: Worker 1 failed executing transaction '9e668a93-2618-11ee-93ee-bc16954181bb:47508257' at master log greatdb-bin.000988, end_log_pos 701570116. See error log and/or performance_schema.replication_applier_status_by_worker table for more details about this failure or others, if any.
                 Skip_Counter: 0
          Exec_Master_Log_Pos: 701548690
              Relay_Log_Space: 2246320360
              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: NULL
Master_SSL_Verify_Server_Cert: No
                Last_IO_Errno: 0
                Last_IO_Error:
               Last_SQL_Errno: 1146
               Last_SQL_Error: Coordinator stopped because there were error(s) in the worker(s). The most recent failure being: Worker 1 failed executing transaction '9e668a93-2618-11ee-93ee-bc16954181bb:47508257' at master log greatdb-bin.000988, end_log_pos 701570116. See error log and/or performance_schema.replication_applier_status_by_worker table for more details about this failure or others, if any.
  Replicate_Ignore_Server_Ids:
             Master_Server_Id: 1943306
                  Master_UUID: 9e668a93-2618-11ee-93ee-bc16954181bb
             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: 230822 14:14:18
               Master_SSL_Crl:
           Master_SSL_Crlpath:
           Retrieved_Gtid_Set: 9e668a93-2618-11ee-93ee-bc16954181bb:2-47565802
            Executed_Gtid_Set: 30873cfe-8750-11ed-b56f-744aa4073024:1-270,
9e668a93-2618-11ee-93ee-bc16954181bb:1-47508256
                Auto_Position: 1
         Replicate_Rewrite_DB:  
                 Channel_Name:
           Master_TLS_Version:
1 row in set (0.00 sec)

根据slave status状态信息可以看出

报错的GTID为：'9e668a93-2618-11ee-93ee-bc16954181bb:47508257'
应用的主集群的binlog为：greatsql-bin.000988
灾备集群的relay log为：greatsql-relay.002963

详细信息查看performance_schema.replication_applier_status_by_worker表

2、查看错误的详细信息

greatsql> select * from performance_schema.replication_applier_status_by_worker\G
*************************** 1. row ***************************
         CHANNEL_NAME:
            WORKER_ID: 1
            THREAD_ID: NULL
        SERVICE_STATE: OFF
LAST_SEEN_TRANSACTION: 9e668a93-2618-11ee-93ee-bc16954181bb:47508257
    LAST_ERROR_NUMBER: 1146
   LAST_ERROR_MESSAGE: Worker 1 failed executing transaction
'9e668a93-2618-11ee-93ee-bc16954181bb:47508257' at master log greatdb-bin.000988,
end_log_pos 701570116; Error executing row event: 'Table 'abs_xxx.tmp_xxx_info' doesn't exist'
LAST_ERROR_TIMESTAMP: 2023-08-22 14:14:18

上述信息说明根据performance_schema.replication_applier_status_by_worker表中的详细错误信息可以发现为灾备集群abs_xxx.tmp_xxx_info表不存在，导致同步报错

3、问题分析

3.1、确认灾备集群中目标表是否存在

greatsql> show create table abs_xxx.tmp_xxx_info;
ERROR 1146 (42S02): Table 'abs_xxx.tmp_xxx_info' doesn't exist
greatsql> desc abs_xxx.tmp_xxx_info;
ERROR 1146 (42S02): Table 'abs_xxx.tmp_xxx_info' doesn't exist

结论：灾备集群中目标表的确不存在

3.2、根据主从报错信息解析主集群binlog，报错的SQL

解析主集群binlog

SET @@SESSION.GTID_NEXT= '9e668a93-2618-11ee-93ee-bc16954181bb:47508257'/*!*/;
……
#230822 14:14:18 server id 1943306  end_log_pos 701570000         Table_map: `abs_xxx`.`tmp_xxx_info` mapped to number 1595
# at 701570000
#230822 14:14:18 server id 1943306  end_log_pos 701570116         Write_rows: table id 1595 flags: STMT_END_F
### INSERT INTO `abs_xxx`.`tmp_xxx_info`
### SET
###   @1=2
###   @2='自动化'
###   @3='2300121212120000'
###   @4='90000000'
###   @5='1'
###   @6='202001290231001'
###   @7='2021-01-31 00:00:00'
# at 701570116
#230822 14:14:18 server id 1943306  end_log_pos 701570143         Xid = 800998400
COMMIT/*!*/;
# at 701570143
#230822 14:14:18 server id 1943306  end_log_pos 701570204         GTID        last_committed=26491        sequence_number=26521        rbr_only=yes
/*!50718 SET TRANSACTION ISOLATION LEVEL READ COMMITTED*//*!*/;

结论：根据复制的报错信息得知具体的GTID号以及主集群的binlog文件，解析binlog得知此事务为一条INSERT语句，语句中的目标表与performance_schema.replication_applier_status_by_worker表中信息一致

3.3、寻找主集群目标表binlog中是否有建表语句

在同一binlog日志中寻找建表语句

SET TIMESTAMP=1692684495/*!*/;
CREATE DATABASE IF NOT EXISTS `abs_xxx` /*!40100 DEFAULT CHARACTER SET utf8 COLLATE utf8_bin */
/*!*/;
……
use `information_schema`/*!*/;
SET TIMESTAMP=1692684495/*!*/;
CREATE TABLE `abs_xxx`.`tmp_xxx_info` (
  `ID` int(64) NOT NULL AUTO_INCREMENT,
  `CUSTOMER_NAME` varchar(256) COLLATE utf8mb4_bin DEFAULT NULL,
  `CUSTOMER_NUM` varchar(256) COLLATE utf8mb4_bin DEFAULT NULL,
  `PRODIST_SKU_NUM` varchar(64) COLLATE utf8mb4_bin DEFAULT NULL,
  `STATUS` varchar(64) COLLATE utf8mb4_bin DEFAULT NULL,
  `AGREEMENT_NUM` varchar(256) COLLATE utf8mb4_bin DEFAULT NULL,
  `END_DATE` datetime DEFAULT NULL,
  PRIMARY KEY (`ID`),
  KEY `PK_PRODIST_SKU_NUM_AGREEMENT` (`PRODIST_SKU_NUM`) USING BTREE,
  KEY `IDX_AGREEMENT_NUM` (`AGREEMENT_NUM`) USING BTREE
) ENGINE=InnoDB AUTO_INCREMENT=2 DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_bin ROW_FORMAT=DYNAMIC
/*!*/;
# at 475864451

结论：在主集群的binlog日志中找到了目标表的建表语句，说明主集群执行DDL时并没有关闭binlog日志，那么继续查看在灾备集群的中继日志中是否存在DDL语句

3.4、解析灾备集群的中继日志，确认是否拉取到灾备集群

#230822 14:08:15 server id 1943306  end_log_pos 475863662         GTID        last_committed=16341        sequence_number=16342        rbr_only=no
SET @@SESSION.GTID_NEXT= '9e668a93-2618-11ee-93ee-bc16954181bb:47498079'/*!*/;
……
use `information_schema`/*!*/;
SET TIMESTAMP=1692684495/*!*/;
CREATE TABLE `abs_xxx`.`tmp_xxx_info` (
  `ID` int(64) NOT NULL AUTO_INCREMENT,
  `CUSTOMER_NAME` varchar(256) COLLATE utf8mb4_bin DEFAULT NULL,
  `CUSTOMER_NUM` varchar(256) COLLATE utf8mb4_bin DEFAULT NULL,
  `PRODIST_SKU_NUM` varchar(64) COLLATE utf8mb4_bin DEFAULT NULL,
  `STATUS` varchar(64) COLLATE utf8mb4_bin DEFAULT NULL,
  `AGREEMENT_NUM` varchar(256) COLLATE utf8mb4_bin DEFAULT NULL,
  `END_DATE` datetime DEFAULT NULL,
  PRIMARY KEY (`ID`),
  KEY `PK_PRODIST_SKU_NUM_AGREEMENT` (`PRODIST_SKU_NUM`) USING BTREE,
  KEY `IDX_AGREEMENT_NUM` (`AGREEMENT_NUM`) USING BTREE
) ENGINE=InnoDB AUTO_INCREMENT=2 DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_bin ROW_FORMAT=DYNAMIC
/*!*/;
# at 475864660
#230822 14:08:15 server id 1943306  end_log_pos 475864512         GTID        last_committed=16342        sequence_number=16343        rbr_only=yes
/*!50718 SET TRANSACTION ISOLATION LEVEL READ COMMITTED*//*!*/;

结论：灾备集群的中继日志中存在DDL建表语句，说明并不是IO线程出了问题

3.5、排查复制配置的忽略库表

Replicate_Ignore_DB: mysql,dbscale,dbscale_tmp,information_schema,performance_schema,sys
Replicate_Wild_Ignore_Table: A.ab,B.bc

结论：忽略库表中并不包含目标表，但是根据以上解析日志发现，在主集群binlog日志中建表语句之前有个use information_schema/!/; 的语句，此库为同步忽略的系统库，因此触发了GreatSQL的规范限制，在忽略库下对未忽略进行操作Statement模式下记录语句默认不起作用（详情：https://dev.mysql.com/doc/refman/5.7/en/replication-options-replica.html#option_mysqld_replicate-do-db）

4、解决同步报错

在灾备集群创建目标表

greatsql> CREATE TABLE `abs_xxx`.`tmp_xxx_info` (
  `ID` int(64) NOT NULL AUTO_INCREMENT,
  `CUSTOMER_NAME` varchar(256) COLLATE utf8mb4_bin DEFAULT NULL,
  `CUSTOMER_NUM` varchar(256) COLLATE utf8mb4_bin DEFAULT NULL,
  `PRODIST_SKU_NUM` varchar(64) COLLATE utf8mb4_bin DEFAULT NULL,
  `STATUS` varchar(64) COLLATE utf8mb4_bin DEFAULT NULL,
  `AGREEMENT_NUM` varchar(256) COLLATE utf8mb4_bin DEFAULT NULL,
  `END_DATE` datetime DEFAULT NULL,
  PRIMARY KEY (`ID`),
  KEY `PK_PRODIST_SKU_NUM_AGREEMENT` (`PRODIST_SKU_NUM`) USING BTREE,
  KEY `IDX_AGREEMENT_NUM` (`AGREEMENT_NUM`) USING BTREE
) ENGINE=InnoDB AUTO_INCREMENT=2 DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_bin ROW_FORMAT=DYNAMIC;

greatsql> stop slave;
greatsql> start slave;

结论：在灾备集群创建目标表后重启复制恢复成功

三、限制规避

1、第一种规避方式

执行DDL时进入目标库

greatsql> use abs_cust
greatsql> DDL 语句（CREATE\DROP\ALTER）

说明：在应用连接数据库时有可能默认就是information_schema库，而此环境将系统库全部忽略，所以为了规避类似的问题，请在执行SQL语句时请先use到目标表的目标库。

2、第二种规避方式

修改主从复制配置，以下步骤为测试环境

关闭灾备集群在复制同步

greatsql> stop slave;
Query OK, 0 rows affected, 1 warning (0.03 sec)

修改忽略库

greatsql> change replication filter Replicate_Ignore_DB=();

修改忽略表

greatsql> change replication filter replicate_wild_ignore_table =('mysql.%','information_schema.%','sys.%','performance_schema.%');

启动同步

greatsql> start slave;
Query OK, 0 rows affected, 1 warning (0.37 sec)

测试验证

主集群：

greatsql> use mysql
Reading table information for completion of table and column names
You can turn off this feature to get a quicker startup with -A
Database changed

greatsql> create table test111.test111(id int primary key);
Query OK, 0 rows affected (0.06 sec)

greatsql> show tables;
+-------------------+
| Tables_in_test111 |
+-------------------+
| test111           |
+-------------------+
1 row in set (0.00 sec)

灾备集群：

greatsql> use test111
Reading table information for completion of table and column names
You can turn off this feature to get a quicker startup with -A
Database changed

greatsql> show tables;
+-------------------+
| Tables_in_test111 |
+-------------------+
| test111           |
+-------------------+
1 row in set (0.00 sec)

说明：复制配置中参数Replicate_Ignore_DB设置为空，将replicate_wild_ignore_table参数设置为shema_name.%的方式也可以规避类似的问题

四、特别说明

在MySQL 5.7跟8.0版本也存在此限制

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：

社区博客有奖征稿详情：https://greatsql.cn/thread-100-1-1.html

技术交流群：

微信：扫码添加GreatSQL社区助手微信好友，发送验证信息加群。