从字面意思看了一下是因为slave_pending_jobs_size_max默认值为16777216(16MB),但是slave接收到的slave_pending_jobs_size_max为17085453(17M);
解决方案
从库执行如下SQL
1
2
3
|
mysql>stop slave;
mysql> set global slave_pending_jobs_size_max=20000000;
mysql> start slave;
|
#在多线程复制时,在队列中Pending的事件所占用的最大内存,默认为16M,如果内存富余,或者延迟较大时,可以适当调大;注意
这个值要比主库的max_allowed_packet大
故障模拟:
从库设置
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
|
mysql> set global slave_parallel_workers=4;
mysql> show variables like 'slave_parallel_workers' ;
+ ------------------------+-------+
| Variable_name | Value |
+ ------------------------+-------+
|slave_parallel_workers | 4 |
+ ------------------------+-------+
1 row in set (0.00sec)
mysql> set global slave_pending_jobs_size_max=1024;
mysql> show variables like 'slave_pend%' ;
+ -----------------------------+-------+
| Variable_name | Value |
+ -----------------------------+-------+
|slave_pending_jobs_size_max | 1024 |
+ -----------------------------+-------+
1 row in set (0.00sec)
|
主库操作:
1
2
3
4
5
6
7
8
9
10
11
|
mysql> update erp_mkpf set usnam= '测试1864错误' ;
ERROR 1197 (HY000):Multi-statement transaction required more than 'max_binlog_cache_size' bytes ofstorage; increase this mysqld variable and try again
mysql> set global max_binlog_cache_size=8388608000000;
Query OK, 0 rowsaffected (0.00 sec)
mysql> begin ;
Query OK, 0 rowsaffected (0.00 sec)
mysql> update erp_mkpf set usnam= '测试1864错误' ;
Query OK, 70466 rowsaffected (0.38 sec)
Rows matched:70466 Changed: 70466 Warnings: 0
mysql> commit ;
Query OK, 0 rowsaffected (0.08 sec)
|
从库查看状态:
1
2
3
|
mysql> show slavestatus\G
Last_SQL_Errno: 1864
Last_SQL_Error: Cannot scheduleevent Update_rows, relay-log name ./HE1-relay-bin.000005, position 494 toWorker thread because its size 8200 exceeds 1024 ofslave_pending_jobs_size_max.
|
以上所述是小编给大家介绍的Mysql 1864 主从错误解决方法,希望对大家有所帮助,如果大家有任何疑问请给我留言,小编会及时回复大家的。在此也非常感谢大家对服务器之家网站的支持!
原文链接:http://suifu.blog.51cto.com/9167728/1841764