First, the data table structure
CREATE TABLE `jx_attach` (
`attach_id` int(11) NOT NULL AUTO_INCREMENT,
`feed_id` int(11) DEFAULT NULL ,
`attach_name` varchar(255) NOT NULL,
`cycore_file_id` varchar(255) DEFAULT NULL ,
`attach_size` bigint(20) NOT NULL DEFAULT '0',
`complete` smallint(6) NOT NULL DEFAULT '0' ,
PRIMARY KEY (`attach_id`),
KEY `jx_trend_attach_FK` (`feed_id`) USING BTREE
) ENGINE=InnoDB AUTO_INCREMENT=394160 DEFAULT CHARSET=utf8;
2. Phenomenon
When multiple connections update the data of a table at the same time, the speed will become slower and slower, and the data table will be locked after a period of time, which will affect other queries and updates.
Stored procedure loops 30 update operations
/*30 update operations*/
BEGIN
DECLARE v1 INT DEFAULT 30;
WHILE v1 > 0 DO
update jx_attach set complete=1,attach_size=63100 where cycore_file_id='56677142da502cd8907eb58f';
SET v1 = v1 - 1;
END WHILE;
END
Execution result (very slow)
Time: 29.876s
Procedure executed successfully
Affected rows: 0
200 data update operations performed simultaneously on three database connections
update jx_attach set complete=1,attach_size=63100 where cycore_file_id='56677142da502cd8907eb58f';
update jx_attach set complete=1,attach_size=63100 where cycore_file_id='56677142da502cd8907eb58f';
update jx_attach set complete=1,attach_size=63100 where cycore_file_id='56677142da502cd8907eb58f';
update jx_attach set complete=1,attach_size=63100 where cycore_file_id='56677142da502cd8907eb58f';
update jx_attach set complete=1,attach_size=63100 where cycore_file_id='56677142da502cd8907eb58f';
update jx_attach set complete=1,attach_size=63100 where cycore_file_id='56677142da502cd8907eb58f';
...etc
Execution result (after a period of time, the speed becomes slower and slower, and there is a waiting lock)
# Time: 151208 22:41:24 # User@Host: zmduan[zmduan] @ [192.168.235.1] Id: 2 # Query_time: 1.848644 Lock_time: 0.780778 Rows_sent: 0 Rows_examined: 393382 SET timestamp=1449643284; update jx_attach set complete=1,attach_size=63100 where cycore_file_id='56677142da502cd8907eb58f'; ......... ........ # User@Host: zmduan[zmduan] @ [192.168.235.1] Id: 2 # Query_time: 2.868598 Lock_time: 1.558542 Rows_sent: 0 Rows_examined: 393382 SET timestamp=1449643805; update jx_attach set complete=1,attach_size=63100 where cycore_file_id='56677142da502cd8907eb58f'; [root@localhost log]# tail -f slow_query.log # User@Host: zmduan[zmduan] @ [192.168.235.1] Id: 19 # Query_time: 1.356797 Lock_time: 0.000169 Rows_sent: 1 Rows_examined: 393383 SET timestamp=1449643805; SELECT * FROM jx_attach ja,jx_feed jf where ja.feed_id=jf.feed_id and ja.cycore_file_id='56677146da502cd8907eb5b7'; # User@Host: zmduan[zmduan] @ [192.168.235.1] Id: 2 # Query_time: 2.868598 Lock_time: 1.558542 Rows_sent: 0 Rows_examined: 393382 SET timestamp=1449643805; update jx_attach set complete=1,attach_size=63100 where cycore_file_id='56677142da502cd8907eb58f';
三、原因分析
MySQL的innodb存储引擎支持行级锁,innodb的行锁是通过给索引项加锁实现的,这就意味着只有通过索引条件检索数据时,innodb才使用行锁,否则使用表锁。根据当前的数据更新语句(update jx_attach set complete=1,attach_size=63100 where cycore_file_id='56677142da502cd8907eb58f';),该条件字段cycore_file_id并没有添加索引,所以导致数据表被锁。
四、解决办法
为cycore_file_id添加索引
五、最终效果(30次更新操作)
时间: 0.094s
Procedure executed successfully
受影响的行: 0