pt-table-sync修復主從庫不一致數據

一、實例演示1

模擬主從庫數據不一致環境:

192.168.0.39 master庫:
mysql> select * from test01.frame01;;
+----+-----------+------------------------------------------+
| id | parent_id | dsn                                      |
+----+-----------+------------------------------------------+
|  1 |         1 | 192.168.0.22,u=pt22,p=ptchecksums,P=3307 |
|  2 |         2 | 192.168.0.33,u=pt33,p=ptchecksums,P=3308 |
+----+-----------+------------------------------------------+
2 rows in set (0.00 sec)

192.168.0.39 slave庫:
mysql> update frame01 set dsn='192.168.0.55,u=umaaa,p=ptchsyeudew,P=3310' where id=1;
Query OK, 1 row affected (0.01 sec)
Rows matched: 1  Changed: 1  Warnings: 0

mysql> update frame01 set dsn='192.168.0.66,u=umbbb,p=ptchsyeudew,P=3311' where id=2;
Query OK, 1 row affected (0.01 sec)
Rows matched: 1  Changed: 1  Warnings: 0
mysql> select * from frame01;
+----+-----------+-------------------------------------------+
| id | parent_id | dsn                                       |
+----+-----------+-------------------------------------------+
|  1 |         1 | 192.168.0.55,u=umaaa,p=ptchsyeudew,P=3310 |
|  2 |         2 | 192.168.0.66,u=umbbb,p=ptchsyeudew,P=3311 |
+----+-----------+-------------------------------------------+

pt-table-sync校驗修復數據具體指令:

[root@kusou-es11 ~]#  /usr/local/percona-toolkit/bin/pt-table-sync h=192.168.0.39,u=ptsum,p=ptchecksums,P=3306 --databases=test01 --tables=frame01 --replicate=percona.checksums --print
REPLACE INTO `test01`.`frame01`(`id`, `parent_id`, `dsn`) VALUES ('1', '1', '192.168.0.22,u=pt22,p=ptchecksums,P=3307') /*percona-toolkit src_db:test01 src_tbl:frame01 src_dsn:P=3306,h=192.168.0.39,p=...,u=ptsum dst_db:test01 dst_tbl:frame01 dst_dsn:P=3306,h=192.168.0.11,p=...,u=ptsum lock:1 transaction:1 changing_src:percona.checksums replicate:percona.checksums bidirectional:0 pid:176411 user:root host:kusou-es11*/;
REPLACE INTO `test01`.`frame01`(`id`, `parent_id`, `dsn`) VALUES ('2', '2', '192.168.0.33,u=pt33,p=ptchecksums,P=3308') /*percona-toolkit src_db:test01 src_tbl:frame01 src_dsn:P=3306,h=192.168.0.39,p=...,u=ptsum dst_db:test01 dst_tbl:frame01 dst_dsn:P=3306,h=192.168.0.11,p=...,u=ptsum lock:1 transaction:1 changing_src:percona.checksums replicate:percona.checksums bidirectional:0 pid:176411 user:root host:kusou-es11*/;

提示: 命令末尾的--print的指令是打印出需要修復數據的命令,不執行命令

pt-table-sync 修復主從庫的表frame01數據 ,使主從庫表frame01數據一致:


[root@kusou-es11 ~]#  /usr/local/percona-toolkit/bin/pt-table-sync h=192.168.0.39,u=ptsum,p=ptchecksums,P=3306 --databases=test01 --tables=frame01 --replicate=percona.checksums --execute
[root@kusou-es11 ~]# 

提示: 命令末尾的 --execute 的指令是執行修復數據的指令,使master庫mysql.user表和slave庫的mysql.user表數據一致

pt-table-checksum 檢測主從庫的表frame01數據一致 (因爲 DIFFS =0 )

[root@kusou-es11 ~]#  /usr/local/percona-toolkit/bin/pt-table-checksum  h=192.168.0.39,u=ptsum,p='ptchecksums',P=3306 --databases=test01 --tables=frame01 --replicate=percona.checksums  --no-check-binlog-format --nocheck-replication-filters  
Checking if all tables can be checksummed ...
Starting checksum ...
            TS ERRORS  DIFFS     ROWS  DIFF_ROWS  CHUNKS SKIPPED    TIME TABLE
06-15T13:41:47      0      0        2          0       1       0   0.316 test01.frame01

slave庫查看,數據和主庫一致

mysql> select * from test01.frame01;
+----+-----------+------------------------------------------+
| id | parent_id | dsn                                      |
+----+-----------+------------------------------------------+
|  1 |         1 | 192.168.0.22,u=pt22,p=ptchecksums,P=3307 |
|  2 |         2 | 192.168.0.33,u=pt33,p=ptchecksums,P=3308 |
+----+-----------+------------------------------------------+
2 rows in set (0.00 sec)

提示:需要注意的是,需要同步的表上必須要有主鍵或者唯一索引,否則會出錯。
同時,pt-table-sync 修復數據時,會造成鎖表,要在業務低峯期來修復主庫的數據

對找到的主從不一致的行,採用replace into語句,在主庫執行一遍以生成該行全量的binlog,並同步到從庫,這會以主庫數據爲基準來修復從庫;
對於主庫有的行而從庫沒有的行,採用replace在主庫上插入(必須不能是×××ert);
對於從庫有而主庫沒有的行,通過在主庫執行delete來刪除(pt-table-sync強烈建議所有的數據修復都只在主庫進行,而不建議直接修改從庫數據;但是也有特例,以後面會講到)。

二、實例演示2

在slave庫上的mysql user表上,新建一個用戶 qdtets@'192.168.0.%' 模擬master和slave庫數據不一致

grant all on *.* to qdtets@'192.168.0.%' identified by 'DHWUOEdwerer';flush privileges;
[root@kusou-es11 ~]#  /usr/local/percona-toolkit/bin/pt-table-checksum  h=192.168.0.39,u=ptsum,p='ptchecksums',P=3306 --databases=mysql --replicate=percona.checksums  --no-check-binlog-format --nocheck-replication-filters  
Checking if all tables can be checksummed ...
Starting checksum ...
            TS ERRORS  DIFFS     ROWS  DIFF_ROWS  CHUNKS SKIPPED    TIME TABLE
06-15T13:32:27      0      0        0          0       1       0   0.317 mysql.columns_priv
06-15T13:32:28      0      0        2          0       1       0   0.318 mysql.db
06-15T13:32:28      0      0        2          0       1       0   0.316 mysql.engine_cost
06-15T13:32:28      0      0        0          0       1       0   0.317 mysql.event
06-15T13:32:29      0      0        0          0       1       0   0.316 mysql.func
06-15T13:32:29      0      0       41          0       1       0   0.317 mysql.help_category
06-15T13:32:29      0      0      699          0       1       0   0.318 mysql.help_keyword
06-15T13:32:30      0      0     1413          0       1       0   0.319 mysql.help_relation
06-15T13:32:30      0      0      643          0       1       0   0.325 mysql.help_topic
06-15T13:32:30      0      0        0          0       1       0   0.316 mysql.ndb_binlog_index
06-15T13:32:31      0      0        0          0       1       0   0.316 mysql.plugin
06-15T13:32:31      0      1       48          0       1       0   0.317 mysql.proc
06-15T13:32:31      0      0        0          0       1       0   0.317 mysql.procs_priv
06-15T13:32:32      0      0        1          0       1       0   0.317 mysql.proxies_priv
06-15T13:32:32      0      0        6          0       1       0   0.317 mysql.server_cost
06-15T13:32:32      0      0        0          0       1       0   0.316 mysql.servers
06-15T13:32:33      0      0        2          0       1       0   0.317 mysql.tables_priv
06-15T13:32:33      0      0        0          0       1       0   0.316 mysql.time_zone
06-15T13:32:33      0      0        0          0       1       0   0.316 mysql.time_zone_leap_second
06-15T13:32:33      0      0        0          0       1       0   0.316 mysql.time_zone_name
06-15T13:32:34      0      0        0          0       1       0   0.318 mysql.time_zone_transition
06-15T13:32:34      0      0        0          0       1       0   0.317 mysql.time_zone_transition_type
**06-15T13:32:34      0      1        6          1       1       0   0.319 mysql.user**

在slave庫上pt-table-sync 修復數據:

slave庫上操作:

[root@kusou-es11 ~]# /usr/local/percona-toolkit/bin/pt-table-sync h=192.168.0.39,u=ptsum,p=ptchecksums,P=3306 --databases=mysql --tables=user --replicate=percona.checksums --print
Access denied for user 'ptsum'@'192.168.0.%' to database 'mysql' [for Statement "LOCK TABLES `mysql`.`user` WRITE"] at line 6172 while doing mysql.user on 192.168.0.11

報錯,提示沒lock tables 權限

解決辦法:登錄master庫重新授權,添加lock tables權限

grant  update,×××ert,select,create,drop,delete,index,execute,lock tables,super,process,replication slave on *.* to ptsum@'192.168.0.%' identified by 'ptchecksums'; flush privileges;

再次操作不在報錯br/>**輸出提示master庫的mysql.user表和slave庫的mysql.user表的數據不一致。slave庫的mysql.user表多了一個用戶qdtets@'192.168.0.%'。需要刪除這個用戶,才能保證主庫和從庫數據的一致
換句話說:
由於從庫只是比主庫多了一條數據,pt-table-sync將以主庫以準,在主庫執行一個刪除操作的事件,然後slave應用此事件完成同步**

[root@kusou-es11 ~]# /usr/local/percona-toolkit/bin/pt-table-sync h=192.168.0.39,u=ptsum,p=ptchecksums,P=3306 --databases=mysql --tables=user --replicate=percona.checksums --print
DELETE FROM `mysql`.`user` WHERE `host`='192.168.0.%' AND `user`='qdtets' LIMIT 1 /*percona-toolkit src_db:mysql src_tbl:user src_dsn:P=3306,h=192.168.0.39,p=...,u=ptsum dst_db:mysql dst_tbl:user dst_dsn:P=3306,h=192.168.0.11,p=...,u=ptsum lock:1 transaction:0 changing_src:percona.checksums replicate:percona.checksums bidirectional:0 pid:137319 user:root host:kusou-es11*/;

提示: 命令末尾的--print的指令是打印出需要修復數據的命令,不執行命令。

使用pt-table-checksum檢測mysql的user表,DIFFS =1 事實證明master和slave的mysql user表數據確實不一致:

[root@kusou-es11 ~]# /usr/local/percona-toolkit/bin/pt-table-checksum  h=192.168.0.39,u=ptsum,p='ptchecksums',P=3306 --databases=mysql --tables=user  --no-check-binlog-format --nocheck-replication-filters
Checking if all tables can be checksummed ...
Starting checksum ...
            TS ERRORS  DIFFS     ROWS  DIFF_ROWS  CHUNKS SKIPPED    TIME TABLE
06-15T09:58:37      0      1        6          1       1       0   0.318 mysql.user

執行修復命令pt-table-sync 修復master和slave的mysql user表數據:

[root@kusou-es11 ~]#  /usr/local/percona-toolkit/bin/pt-table-sync h=192.168.0.39,u=ptsum,p=ptchecksums,P=3306 --databases=mysql --tables=user --replicate=percona.checksums --execute
[root@kusou-es11 ~]# 

提示: 命令末尾的 --execute 的指令是執行修復數據的指令,使master庫mysql.user表和slave庫的mysql.user表數據一致

slave庫上檢查master庫的mysql.user表 和slave庫的mysql.user 表數據是否一致:

[root@kusou-es11 ~]# /usr/local/percona-toolkit/bin/pt-table-checksum  h=192.168.0.39,u=ptsum,p='ptchecksums',P=3306 --databases=mysql --tables=user  --no-check-binlog-format --nocheck-replication-filters
Checking if all tables can be checksummed ...
Starting checksum ...
            TS ERRORS  DIFFS     ROWS  DIFF_ROWS  CHUNKS SKIPPED    TIME TABLE
06-15T09:46:52      0      0        6          0       1       0   0.319 mysql.user

可以看到DIFFS =0 ,說明數據已經修復完成

三、實例演示3

將master 上的所有數據同步到slave:

[root@kusou-es11 ~]# /usr/local/percona-toolkit/bin/pt-table-sync --execute  h=192.168.0.39,u=ptsum,p=ptchecksums,P=3306 --databases=mysql --tables=user h=192.168.0.11,u=ptsum,p='ptchecksums',P=3306 --no-check-slave  --print
DELETE FROM `mysql`.`user` WHERE `host`='10.0.0.1' AND `user`='dhrue' LIMIT 1 /*percona-toolkit src_db:mysql src_tbl:user src_dsn:P=3306,h=192.168.0.39,p=...,u=ptsum dst_db:mysql dst_tbl:user dst_dsn:P=3306,h=192.168.0.11,p=...,u=ptsum lock:0 transaction:0 changing_src:0 replicate:0 bidirectional:0 pid:12565 user:root host:kusou-es11*/;

將master 上的所有數據同步到slave1和slave2:


[root@kusou-es11 ~]# /usr/local/percona-toolkit/bin/pt-table-sync --execute  h=192.168.0.39,u=ptsum,p=ptchecksums,P=3306 --databases=mysql --tables=user h=192.168.0.11,u=ptsum,p='ptchecksums',P=3307    h=192.168.0.22,u=ptsum,p='ptchecksums',P=3308   --no-check-slave  --print

對pt-table-checksum和pt-table-sync這一組工具進行了最簡單的測試,其實運行這一組命令不一定需要在主從結構的主庫上進行,網段內的任何服務器都可以運行,前提就是安裝好這套工具就好。

四、原理介紹

pt-table-checksum/pt-table-sync原理介紹可以參考如下博文:
http://blog.sina.com.cn/s/blog_a1e9c7910102vnsd.html

發表評論
所有評論
還沒有人評論,想成為第一個評論的人麼? 請在上方評論欄輸入並且點擊發布.
相關文章