1.认真看错误提示,弄不清楚状况的时候打开debugPTDEBUG=1
2.10.10.10.10为主,其它ip为从
3.h=10.10.10.10,u=checksums,p=checksums 中间一定不能有空格,不然会报错,本人排查这个错误花了很长时间
一:sakila库的film_actor表的数据检测与同步
1,在主库和从库上面添加用户:
GRANT SELECT, PROCESS, SUPER, REPLICATION SLAVE ON *.* TO ’checksums’@’10.10.10.10’ IDENTIFIED BY ’checksums’;
2.检测sakila库的film_actor表的数据主从是否一致
PTDEBUG=1 pt-table-checksum –no-check-binlog-format –no-check-replication-filters –recursion-method=processlist –replicate=test.checksums –databases=sakila –tables=film_actor -h 10.10.10.10 -P8888 -u checksums -p checksums
3,根据checksum的结果恢复所有从库的film_actor数据
PTDEBUG=1 pt-table-sync –replicate=test.checksums –recursion-method=processlist –database=sakila –tables=film_actor h=10.10.10.10,P=8888,u=checksums,p=checksums –print –execute
——————————————————————————————————————————
二:检测tmp库数据是否一致
pt-table-checksum –no-check-binlog-format –no-check-replication-filters –recursion-method=processlist –replicate=test.checksums –databases=tmp -h 10.10.10.10 -P8888 -u checksums -p checksums
1,根据checksum的结果恢复所有从库的数据
PTDEBUG=1 pt-table-sync –replicate=test.checksums –recursion-method=processlist –no-check-triggers –no-check-child-tables –no-foreign-key-checks h=10.10.10.10,P=8888,u=checksums,p=checksums –print –execute
(有外键约束要加参数 –no-check-child-tables –no-foreign-key-checks,不然可能把子表所有的数据都删除掉)
(–no-check-triggers,不检测触发器)
2,根据checksum的结果只恢复10.10.10.11的数据
pt-table-sync –print –execute –replicate=test.checksums –sync-to-master h=10.10.10.11,P=8888,u=checksums,p=checksums
3,恢复从库192.168.6.91的数据(这个从库所有的数据都和主保持一致,除了系统表和checksums表)
pt-table-sync –execute –print –sync-to-master h=192.168.6.91,P=8888,u=checksums,p=checksums
—————————————————————————————————————————-
执行完pt-table-sync后可再执行一次pt-table-checksum检测主从数据是否一致,可以执行下面的sql语句,如果结果为空则说明主从数据一致
select * from test.checksums where master_cnt <>this_cnt OR master_crc<>this_crc OR ISNULL(master_crc)<>ISNULL(this_crc);
http://blog.itpub.net/28939273/viewspace-1993010/