bitsCN.com
主从(master/slave)默认引擎不一致导致复制失败解决
同事告知，一台slave服务器复制中断，查看slave服务器,如下错误
Error 'Specified key was too long; max keylength is 1000 bytes' on query
mysql> show slave status /G;
*************************** 1. row***************************
Slave_IO_State: Waiting formaster to send event
Master_Host: 192.xxx.xxx.146
Master_User: repl
Master_Port: 3306
Connect_Retry: 60
Master_Log_File: mysql-bin.004369
Read_Master_Log_Pos: 717957570
Relay_Log_File: relay-bin.013649
Relay_Log_Pos: 571890895
Relay_Master_Log_File: mysql-bin.004368
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: 1071
Last_Error: Error 'Specified key was too long; max key length is 1000 bytes' onquery. Default database: 'analyse_xxx'. Query: 'CREATE TABLE`meta_topic_scan` ( `domain` varchar(200) NOT NULL COMMENT '' , `topic_name`varchar(200) NOT NULL COMMENT '' , `topic_url` varchar(200) NOT NULL COMMENT '', `topic_pv` int NULL DEFAULT 0 COMMENT '' , `topic_uv` int NULL DEFAULT 0COMMENT '' , `pv` int NULL DEFAULT 0 COMMENT '' , `uv` int NULL DEFAULT 0COMMENT '' , `shopcart_pv` int NULL DEFAULT 0 COMMENT ''
Skip_Counter: 0
Exec_Master_Log_Pos: 571890750
Relay_Log_Space: 1802530754
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: 1071
Last_SQL_Error: Error 'Specified keywas too long; max key length is 1000 bytes' on query. Default database:'analyse_product'. Query: 'CREATE TABLE
`meta_topic_scan` ( `domain` varchar(200)NOT NULL COMMENT '' , `topic_name` varchar(200) NOT NULL COMMENT '' ,`topic_url` varchar(200) NOT NULL COMMENT '' , `topic_pv` int NULL DEFAULT 0COMMENT '' , `topic_uv` int NULL DEFAULT 0 COMMENT '' , `pv` int NULL DEFAULT 0COMMENT '' , `uv` int NULL DEFAULT 0 COMMENT '' , `shopcart_pv` int NULL DEFAULT0
1 row in set (0.00 sec)
ERROR:
No query specified
创建表竟然失败,索引长度超过了最大值1000，看看报错日志
[root@db~]# tail -n 500 /my/data/mysql/dbxxxxxx.err
130606 10:57:20 [ERROR] Slave SQL: Error'Specified key was too long; max key length is 1000 bytes' on query. Defaultdatabase: 'analyse_xxx'. Query: 'CREATE TABLE
`meta_topic_scan` ( `domain` varchar(200)NOT NULL COMMENT '' , `topic_name` varchar(200) NOT NULL COMMENT '' ,`topic_url` varchar(200) NOT NULL COMMENT '' , `topic_pv` int
NULL DEFAULT 0 COMMENT '' , `topic_uv` intNULL DEFAULT 0 COMMENT '' , `pv` int NULL DEFAULT 0 COMMENT '' , `uv` int NULLDEFAULT 0 COMMENT '' , `shopcart_pv` int NULL DEFAULT 0
130606 10:57:20 [Warning] Slave: Specifiedkey was too long; max key length is 1000 bytes Error_code: 1071
130606 10:57:20 [ERROR] Error runningquery, slave SQL thread aborted. Fix the problem, and restart the slave SQLthread with "SLAVE START". We stopped at log 'mysql-bin.004368' position571890750
日志显示在上午10点57分建表失败，很明显的错误，索引长度超过了1000字节，然后告知同事，创建表索引长度超过了1000字节，导致slave io_thread中断。然后同事很奇怪的说，上午master上面创建表没问题，顺利的创建，为什么在slave上复制报错？
后来通过在slave上手工建表发现master和slave 默认存储引擎不一致，不一致会引起中断？是的，建表语句是否明确指定使用那种存储引擎如下语句
CREATE TABLE `meta_topic_scan` (
`domain` varchar(200) NOT NULL ,
`topic_name` varchar(200) NOT NULL,
`topic_url` varchar(200) NOT NULL ,
`topic_pv` int(11) DEFAULT '0',
`topic_uv` int(11) DEFAULT '0',
PRIMARY KEY (`domain`,`topic_name`,`topic_url`)
);
如果在建表时不指定存储引擎，在master上默认引擎是innodb，创建的表就是innodb类型，写到binlog日志里的语句是什么格式呢？
我们利用mysqlbinlog去查找这条语句，如下格式去分析
[root@aeolus1 ~]# mysqlbinlog -uroot--start-position=1914 /usr/local/mysql/data/mysql-bin.000014
发现建表语句以这样的格式写入binlog二进制日志中
CREATE TABLE `meta_topic_scan` (
`domain` varchar(200) NOT NULL comment '',
`topic_name` varchar(200) NOT NULL comment '',
`topic_url` varchar(200) NOT NULL comment '',
`topic_pv` int(11) DEFAULT '0' comment '',
`topic_uv` int(11) DEFAULT '0' comment '',
PRIMARY KEY (`domain`,`topic_name`,`topic_url`)
);
这样io_thread把binlog从master服务器传送到slave，sql_thread在slave上执行此语句，由此可知slave默认是什么引擎，就创建默认引擎的表。由于slave默认是mysiam，索引在innodb和mysiam两个引擎上索引长度不一样，导致主从复制中断。
修改my.cnf，改默认引擎innodb，重启服务(业务不是很重要，允许重启)，ok。如果不允许重启，那就在slave上创建表，然后跳过出错的event，在维护窗口再重启生效
bitsCN.com
