背景：
一個簡單的主從結構，主的binlog format是Mixed模式，在執行一條簡單的導入語句時，通過mysqlbinlog導出發現記錄的Binlog全部變成了Row的格式（明明設置的是Mixed），現在就說說在什么情況下Binlog的Mixed模式會轉變記錄成Row格式。
概念：
binlog format有三種形式：Statement、Mixed、Row，具體的信息可以自行到網上搜查。
分析（本文碰到的案例）：
查看MySQL binlog format
dba@192.168.111.4 : dba_test 02:33:39>show variables like 'binlog_format%'; +---------------+-------+ | Variable_name | Value | +---------------+-------+ | binlog_format | MIXED | +---------------+-------+
測試語句：
dba@192.168.111.4 : dba_test 02:24:14>create table tmp_test(id int,name varchar(64),age int,primary key(id)) engine = innodb; Query OK, 0 rows affected (0.05 sec) dba@192.168.111.4 : dba_test 02:24:23>insert into tmp_test values(1,'aaa',11); Query OK, 1 row affected (0.02 sec) dba@192.168.111.4 : dba_test 02:25:17>insert into tmp_test values(2,'bbb',22); Query OK, 1 row affected (0.02 sec) dba@192.168.111.4 : dba_test 02:25:23>insert into tmp_test values(3,'ccc',33); Query OK, 1 row affected (0.01 sec) dba@192.168.111.4 : dba_test 02:25:28>insert into tmp_test values(4,'ddd',44); Query OK, 1 row affected (0.01 sec) dba@192.168.111.4 : dba_test 02:25:34>insert into tmp_test values(5,'eee',55); Query OK, 1 row affected (0.01 sec) dba@192.168.111.4 : dba_test 02:25:42>select * from tmp_test; +----+------+------+ | id | name | age | +----+------+------+ | 1 | aaa | 11 | | 2 | bbb | 22 | | 3 | ccc | 33 | | 4 | ddd | 44 | | 5 | eee | 55 | +----+------+------+ 5 rows in set (0.01 sec) dba@192.168.111.4 : dba_test 02:25:50>create table tmp_test_bak(id int,name varchar(64),age int,primary key(id)) engine = innodb; Query OK, 0 rows affected (0.03 sec) dba@192.168.111.4 : dba_test 02:26:31>insert into tmp_test_bak select * from tmp_test; ###記錄成了Row模式 Query OK, 5 rows affected (0.03 sec) Records: 5 Duplicates: 0 Warnings: 0
Binlog 記錄圖：
問題來了，我想要出來的binlog format是Statement，而不是Row。而一條insert into tb select * from ta的簡單語句在Mixed模式下記錄了Row模式的binlog。原因是什么？
首先確實在一些特定的情況下，Mixed會被轉換成Row模式：
. 當 DML 語句更新一個 NDB 表時； . 當函數中包含 UUID() 時； . 2 個及以上包含 AUTO_INCREMENT 字段的表被更新時； . 執行 INSERT DELAYED 語句時； . 用 UDF 時； . 視圖中必須要求運用 row 時，例如建立視圖時使用了 UUID() 函數；
上面來自網絡，有興趣的可以自己測試測試。而對於本文中的sql，符合不了上面的條件，但binlog也記錄成了Row格式。所以還是很奇怪為什么binlog格式被轉換了，日常工作的時候有遇到過執行一條sql，會報一個warning：
Warning: Unsafe statement written to the binary log using statement format since BINLOG_FORMAT = STATEMENT
難道因為這個導致轉換的？因為上面的SQL可以重現，沒有報warning，所以這個情況排除。根據經驗想到了一個參數：innodb_locks_unsafe_for_binlog，看到里面講到事務隔離級別，那就看看隔離級別的情況：
dba@192.168.111.4 : dba_test 05:46:56>select @@global.tx_isolation; +-----------------------+ | @@global.tx_isolation | +-----------------------+ | READ-COMMITTED | +-----------------------+ 1 row in set (0.01 sec) dba@192.168.111.4 : dba_test 06:36:45>select @@session.tx_isolation; +------------------------+ | @@session.tx_isolation | +------------------------+ | READ-COMMITTED | +------------------------+ 1 row in set (0.01 sec)
看到隔離級別是提交讀，即不可重復讀。把事務隔離級別設置成默認的 REPEATABLE READ：
dba@192.168.111.4 : dba_test 06:41:02>set session transaction isolation level REPEATABLE READ; Query OK, 0 rows affected (0.14 sec) dba@192.168.111.4 : dba_test 06:41:42>select @@session.tx_isolation; +------------------------+ | @@session.tx_isolation | +------------------------+ | REPEATABLE-READ | +------------------------+ 1 row in set (0.00 sec)
再執行測試里的SQL，發現這時候Mixed的binlog記錄了Statement格式，正常了，符合預期了。難道就是這個事務隔離級別的問題引起的？在手冊里發現了這句：
Note
In MySQL 5.7, when READ COMMITTED isolation level is used, or the deprecated innodb_locks_unsafe_for_binlog system variable is enabled,
there is no InnoDB gap locking except for foreign-key constraint checking and duplicate-key checking. Also, record locks for nonmatching
rows are released after MySQL has evaluated the WHERE condition.
If you use READ COMMITTED or enable innodb_locks_unsafe_for_binlog, you must use row-based binary logging.
展開可以看例子：
dba@192.168.111.4 : dba_test 06:49:27>select @@session.tx_isolation; +------------------------+ | @@session.tx_isolation | +------------------------+ | REPEATABLE-READ | +------------------------+ 1 row in set (0.01 sec) dba@192.168.111.4 : dba_test 06:49:29>create table tmp_test_bak(id int,name varchar(64),age int,primary key(id)) engine = innodb; Query OK, 0 rows affected (0.14 sec) dba@192.168.111.4 : dba_test 06:49:34>insert into tmp_test_bak values(1,'aaa',11); Query OK, 1 row affected (0.13 sec) dba@192.168.111.4 : dba_test 06:49:41>insert into tmp_test_bak values(2,'bbb',22); Query OK, 1 row affected (0.11 sec) dba@192.168.111.4 : dba_test 06:49:41>insert into tmp_test_bak values(3,'ccc',33); Query OK, 1 row affected (0.11 sec) dba@192.168.111.4 : dba_test 06:49:42>insert into tmp_test_bak values(4,'ddd',44); Query OK, 1 row affected (0.11 sec) dba@192.168.111.4 : dba_test 06:49:42>insert into tmp_test_bak values(5,'eee',55); Query OK, 1 row affected (0.13 sec) BINLOG: root@dba:/var/log/mysql# mysqlbinlog mysql-bin.000007 |grep "tmp_test_bak" create table tmp_test_bak(id int,name varchar(64),age int,primary key(id)) engine = innodb insert into tmp_test_bak values(1,'aaa',11) insert into tmp_test_bak values(2,'bbb',22) insert into tmp_test_bak values(3,'ccc',33) insert into tmp_test_bak values(4,'ddd',44) insert into tmp_test_bak values(5,'eee',55) ############################# ############################# dba@192.168.111.4 : dba_test 06:49:43>set session transaction isolation level READ COMMITTED; Query OK, 0 rows affected (0.01 sec) dba@192.168.111.4 : dba_test 06:50:21>flush logs; Query OK, 0 rows affected (0.21 sec) dba@192.168.111.4 : dba_test 06:50:23>create table tmp_test(id int,name varchar(64),age int,primary key(id)) engine = innodb; Query OK, 0 rows affected (0.14 sec) dba@192.168.111.4 : dba_test 06:50:49>insert into tmp_test values(1,'aaa',11); Query OK, 1 row affected (0.13 sec) dba@192.168.111.4 : dba_test 06:50:54>insert into tmp_test values(2,'bbb',22); Query OK, 1 row affected (0.11 sec) dba@192.168.111.4 : dba_test 06:50:54>insert into tmp_test values(3,'ccc',33); Query OK, 1 row affected (0.11 sec) dba@192.168.111.4 : dba_test 06:50:54>insert into tmp_test values(4,'ddd',44); Query OK, 1 row affected (0.11 sec) dba@192.168.111.4 : dba_test 06:50:54>insert into tmp_test values(5,'eee',55); Query OK, 1 row affected (0.12 sec) BINLOG: root@pubmed2:/var/log/mysql# mysqlbinlog -vv mysql-bin.000008 |grep "tmp_test" create table tmp_test(id int,name varchar(64),age int,primary key(id)) engine = innodb #160426 18:50:54 server id 1 end_log_pos 395 Table_map: `pubmed_data`.`tmp_test` mapped to number 44 ### INSERT INTO `pubmed_data`.`tmp_test` #160426 18:50:54 server id 1 end_log_pos 597 Table_map: `pubmed_data`.`tmp_test` mapped to number 44 ### INSERT INTO `pubmed_data`.`tmp_test` #160426 18:50:54 server id 1 end_log_pos 799 Table_map: `pubmed_data`.`tmp_test` mapped to number 44 ### INSERT INTO `pubmed_data`.`tmp_test` #160426 18:50:54 server id 1 end_log_pos 1001 Table_map: `pubmed_data`.`tmp_test` mapped to number 44 ### INSERT INTO `pubmed_data`.`tmp_test` #160426 18:50:55 server id 1 end_log_pos 1203 Table_map: `pubmed_data`.`tmp_test` mapped to number 44 ### INSERT INTO `pubmed_data`.`tmp_test`
經過測試，在5.1、5.5、5.6都有這個情況，可能這個本身就不是問題。:)
總結：
除了上面說的一些條件影響binlog的記錄格式外（Statement 到 Row），事務隔離也影響這binlog的記錄格式行為。