pt-table-checksum检测不出主从差异处理

Posted 2020-11-03 醒嘞

几个月前写过pt-table-checksum 3.0.4检测不出主从差异数据，当时的解决方案是使用旧版本，另一个挫方法是自行设置binlog_format=\'STATEMENT\'。现在已经发布到3.0.9版本，结果又遇到相同的坑~
最近几版pt-table-checksum(已核实3.0.4和3.0.9有问题)在binlog_format=\'row\'，且主从存在差异数据时，却检测不出主从差异。原因就是处理过程中主上没有SET @@binlog_format := \'STATEMENT\'，导致下面两个核心语句不是以statement格式记录，从库不会进行CRC32相关运算，主从永远一致~

# pt-table-checksum 3.0.4检测不出差异数据
[root@ZST2 ~]# /usr/local/bin/pt-table-checksum --nocheck-binlog-format --nocheck-replication-filters --recursion-method=hosts --replicate=replcrash.checksums --databases=replcrash --tables=py_user,py_user_innodb --host=192.168.85.132 --port=3306 --user=mydba --password=mysql5721
            TS ERRORS  DIFFS     ROWS  CHUNKS SKIPPED    TIME TABLE
05-08T09:41:15      0      0        8       1       0   0.257 replcrash.py_user
05-08T09:41:16      0      0    67740       5       0   1.056 replcrash.py_user_innodb

# 两个核心语句
REPLACE INTO `replcrash`.`checksums` (db, tbl, chunk, chunk_index, lower_boundary, upper_boundary, this_cnt, this_crc) SELECT \'replcrash\', \'py_user\', \'1\', NULL, NULL, NULL, COUNT(*) AS cnt, COALESCE(LOWER(CONV(BIT_XOR(CAST(CRC32(CONCAT_WS(\'#\', `uid`, convert(`name` using utf8mb4), `add_time`, convert(`server_id` using utf8mb4), CONCAT(ISNULL(`name`), ISNULL(`add_time`), ISNULL(`server_id`)))) AS UNSIGNED)), 10, 16)), 0) AS crc FROM `replcrash`.`py_user` /*checksum table*/
UPDATE `replcrash`.`checksums` SET chunk_time = \'0.004092\', master_crc = \'5abbd632\', master_cnt = \'8\' WHERE db = \'replcrash\' AND tbl = \'py_user\' AND chunk = \'1\'

View Code

通过对比旧版本(2.2.3)的/usr/local/bin/pt-table-checksum代码，发现只需稍微调整代码即可正常使用

# pt-table-checksum 3.0.4 只需注释掉第9335行和9364行，这层逻辑应该应用于任何情况下，因此不需要使用if判断
[root@ZST2 ~]# vim /usr/local/bin/pt-table-checksum
...
9335       #if ( $o->get(\'check-binlog-format\') ) {
9336         # https://bugs.launchpad.net/percona-toolkit/+bug/919352
9337         # The tool shouldn\'t blindly attempt to change binlog_format;
9338         # instead, it should check if it\'s already set to STATEMENT.
9339         # This is becase starting with MySQL 5.1.29, changing the format
9340         # requires a SUPER user.
9341         if ( VersionParser->new($dbh) >= \'5.1.5\' ) {
9342            $sql = \'SELECT @@binlog_format\';
9343            PTDEBUG && _d($dbh, $sql);
9344            my ($original_binlog_format) = $dbh->selectrow_array($sql);
9345            PTDEBUG && _d(\'Original binlog_format:\', $original_binlog_format);
9346            if ( $original_binlog_format !~ /STATEMENT/i ) {
9347               $sql = q{/*!50108 SET @@binlog_format := \'STATEMENT\'*/};
9348               eval {
9349                  PTDEBUG && _d($dbh, $sql);
9350                  $dbh->do($sql);
9351               };
9352               if ( $EVAL_ERROR ) {
9353                  die "Failed to $sql: $EVAL_ERROR\\n"
9354                     . "This tool requires binlog_format=STATEMENT, "
9355                     . "but the current binlog_format is set to "
9356                     ."$original_binlog_format and an error occurred while "
9357                     . "attempting to change it.  If running MySQL 5.1.29 or newer, "
9358                     . "setting binlog_format requires the SUPER privilege.  "
9359                     . "You will need to manually set binlog_format to \'STATEMENT\' "
9360                     . "before running this tool.\\n";
9361               }
9362            }
9363         }
9364       #}
...

# 查看pt版本
[root@ZST2 ~]# /usr/local/bin/pt-table-checksum --version
pt-table-checksum 3.0.4
[root@ZST2 ~]# 
# 修改后检测出主从不一致
[root@ZST2 ~]# /usr/local/bin/pt-table-checksum --nocheck-binlog-format --nocheck-replication-filters --recursion-method=hosts --replicate=replcrash.checksums --databases=replcrash --tables=py_user,py_user_innodb --host=192.168.85.132 --port=3306 --user=mydba --password=mysql5721
            TS ERRORS  DIFFS     ROWS  CHUNKS SKIPPED    TIME TABLE
05-08T09:51:32      0      1        8       1       0   0.030 replcrash.py_user
05-08T09:51:33      0      0    67740       5       0   0.561 replcrash.py_user_innodb
[root@ZST2 ~]# 

# general-log信息
[root@ZST1 ~]# cat /data/mysql/mysql3306/data/ZST1.log |more
...
2018-05-08T01:51:32.315064Z        30 Query     SHOW VARIABLES LIKE \'version%\'
2018-05-08T01:51:32.318337Z        30 Query     SHOW ENGINES
2018-05-08T01:51:32.319229Z        30 Query     SHOW VARIABLES LIKE \'innodb_version\'
2018-05-08T01:51:32.322518Z        30 Query     SELECT @@binlog_format
2018-05-08T01:51:32.323019Z        30 Query     /*!50108 SET @@binlog_format := \'STATEMENT\'*/
2018-05-08T01:51:32.323453Z        30 Query     SET SESSION TRANSACTION ISOLATION LEVEL REPEATABLE READ
2018-05-08T01:51:32.324012Z        30 Query     SHOW VARIABLES LIKE \'wsrep_on\'
2018-05-08T01:51:32.327830Z        30 Query     SELECT @@SERVER_ID
2018-05-08T01:51:32.328416Z        30 Query     SHOW SLAVE HOSTS
2018-05-08T01:51:32.344178Z        30 Query     SHOW VARIABLES LIKE \'wsrep_on\'
2018-05-08T01:51:32.347571Z        30 Query     SELECT @@SERVER_ID
2018-05-08T01:51:32.351692Z        30 Query     SHOW VARIABLES LIKE \'wsrep_on\'
2018-05-08T01:51:32.355034Z        30 Query     SELECT @@SERVER_ID
2018-05-08T01:51:32.359543Z        30 Query     SHOW DATABASES LIKE \'replcrash\'
2018-05-08T01:51:32.360406Z        30 Query     CREATE DATABASE IF NOT EXISTS `replcrash` /* pt-table-checksum */
2018-05-08T01:51:32.361538Z        30 Query     USE `replcrash`
2018-05-08T01:51:32.362069Z        30 Query     SHOW TABLES FROM `replcrash` LIKE \'checksums\'
2018-05-08T01:51:32.364738Z        30 Query     CREATE TABLE IF NOT EXISTS `replcrash`.`checksums` (
     db             CHAR(64)     NOT NULL,
     tbl            CHAR(64)     NOT NULL,
     chunk          INT          NOT NULL,
     chunk_time     FLOAT            NULL,
     chunk_index    VARCHAR(200)     NULL,
     lower_boundary TEXT             NULL,
     upper_boundary TEXT             NULL,
     this_crc       CHAR(40)     NOT NULL,
     this_cnt       INT          NOT NULL,
     master_crc     CHAR(40)         NULL,
     master_cnt     INT              NULL,
     ts             TIMESTAMP    NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
     PRIMARY KEY (db, tbl, chunk),
     INDEX ts_db_tbl (ts, db, tbl)
  ) ENGINE=InnoDB DEFAULT CHARSET=utf8
2018-05-08T01:51:32.920625Z        30 Query     SHOW GLOBAL STATUS LIKE \'Threads_running\'
2018-05-08T01:51:32.924636Z        30 Query     SELECT CONCAT(@@hostname, @@port)
...

View Code

如果遇到类似问题，建议开启general_log，查看处理过程，再核验到底什么原因导致检测不出主从差异数据(･ω･)

---

09:21 2018/5/10 补充，其实可以在命令行带上--set-vars binlog_format=\'statement\'
详细说明参考：pt-table-checksum not detecting diffs

[root@ZST2 ~]# /usr/local/bin/pt-table-checksum --nocheck-binlog-format --nocheck-replication-filters --recursion-method=hosts --replicate=replcrash.checksums --databases=replcrash --tables=py_user,py_user_innodb --host=192.168.85.132 --port=3306 --user=mydba --password=mysql5721 --set-vars binlog_format=\'statement\'
            TS ERRORS  DIFFS     ROWS  CHUNKS SKIPPED    TIME TABLE
05-10T09:25:01      0      1       11       1       0   0.024 replcrash.py_user
05-10T09:25:01      0      0    67740       5       0   0.695 replcrash.py_user_innodb
[root@ZST2 ~]# 

View Code

善于思考，善于搜索，不要期盼别人把结果原封不动地送到嘴里~.~