mysql複製和記憶體引擎的表

清晰的格式：
https://zhuanlan.zhihu.com/p/26985377

When a master server shuts down and restarts, its MEMORY tables become empty. To replicate this effect to slaves, the first time that the master uses a given MEMORY table after startup, it logs an event that notifies slaves that the table must to be emptied by writing a DELETE statement for that table to the binary log.

When a slave server shuts down and restarts, its MEMORY tables become empty. This causes the slave to be out of synchrony with the master and may lead to other failures or cause the slave to stop:

Row-format updates and deletes received from the master may fail with Can't find record in 'memory_table'.
Statements such as INSERT INTO ... SELECT FROM memory_table may insert a different set of rows on the master and slave.

當主庫上使用記憶體引擎的表時，主庫因為某些原因重啟了。啟動好之後，再開啟主庫的記憶體引擎表時，會先把裡面的內容清空。所以這個時候主從狀態是正常的。

但當從庫被重啟時，因為memory的內容沒有了，所以可能會出現主從報錯的現象。

實驗：

一主庫異常關閉：

在主庫上建表

mysql> show create table b;

+-------+-------------------------------------------------------------------------------------------------------+

| Table | Create Table |

+-------+-------------------------------------------------------------------------------------------------------+

| b | CREATE TABLE `b` (

`id` int(11) NOT NULL,

PRIMARY KEY (`id`)

) ENGINE=MEMORY DEFAULT CHARSET=utf8

主庫：

mysql> insert into b values (10),(20);

Query OK, 2 rows affected (0.00 sec)

Records: 2 Duplicates: 0 Warnings: 0

從庫：

mysql> select * from a.b;

+----+

| id |

+----+

| 10 |

| 20 |

+----+

重啟主庫後：

從庫：

mysql> select * from a.b;

+----+

| id |

+----+

| 10 |

| 20 |

+----+

2 rows in set (0.00 sec)

這時再去主庫上觀察：

mysql> show master status;

+------------------+----------+--------------+------------------+-------------------------------------------+

+------------------+----------+--------------+------------------+-------------------------------------------+

+------------------+----------+--------------+------------------+-------------------------------------------+

1 row in set (0.00 sec)

mysql> select *from a.b;

Empty set (0.00 sec)

mysql> show master status;

+------------------+----------+--------------+------------------+-------------------------------------------+

+------------------+----------+--------------+------------------+-------------------------------------------+

+------------------+----------+--------------+------------------+-------------------------------------------+

1 row in set (0.00 sec)

mysql> show binlog events in 'mysql-bin.000012';

+------------------+-----+----------------+-----------+-------------+--------------------------------------------------------------------+

+------------------+-----+----------------+-----------+-------------+--------------------------------------------------------------------+

| mysql-bin.000012 | 4 | Format_desc | 169454186 | 123 | Server ver: 5.7.16.k1-ucloudrel1-log, Binlog ver: 4 |

| mysql-bin.000012 | 123 | Previous_gtids | 169454186 | 194 | 48dfe7f5-3ab5-11e7-b3fa-525400199b09:1-23 |

| mysql-bin.000012 | 194 | Gtid | 169454186 | 259 | SET @@SESSION.GTID_NEXT= '48dfe7f5-3ab5-11e7-b3fa-525400199b09:24' |

| mysql-bin.000012 | 259 | Query | 169454186 | 327 | BEGIN |

| mysql-bin.000012 | 327 | Query | 169454186 | 409 | DELETE FROM `a`.`b` |

| mysql-bin.000012 | 409 | Query | 169454186 | 478 | COMMIT |

+------------------+-----+----------------+-----------+-------------+--------------------------------------------------------------------+

6 rows in set (0.00 sec)

再次去從庫觀察：

mysql> select * from a.b;

Empty set (0.00 sec)

這時主從的狀態也是正常的。

二從庫啟常重啟：

主庫上：

mysql> insert into a.b values (10),(20);

Query OK, 2 rows affected (0.00 sec)

Records: 2 Duplicates: 0 Warnings: 0

從庫：

mysql> select * from a.b;

+----+

| id |

+----+

| 10 |

| 20 |

+----+

2 rows in set (0.00 sec)

重啟從庫後：

mysql> select *from a.b;

Empty set (10.00 sec)

主庫：

mysql> delete from a.b where id=10;

Query OK, 1 row affected (0.00 sec)

從庫：

mysql> show slave status \G

*************************** 1. row ***************************

Slave_IO_State: Waiting for master to send event

Master_Host: 10.25.170.106

Master_User: ucloudbackup

Master_Port: 3306

Connect_Retry: 60

Master_Log_File: mysql-bin.000012

Read_Master_Log_Pos: 1049

Relay_Log_File: mysql-relay.000024

Relay_Log_Pos: 360

Relay_Master_Log_File: mysql-bin.000012

Slave_IO_Running: Yes

Slave_SQL_Running: No

Replicate_Do_DB:

Replicate_Ignore_DB:

Replicate_Do_Table:

Replicate_Ignore_Table:

Replicate_Wild_Do_Table:

Replicate_Wild_Ignore_Table:

Last_Errno: 1032

Last_Error: Coordinator stopped because there were error(s) in the worker(s). The most recent failure being: Worker 0 failed executing transaction '48dfe7f5-3ab5-11e7-b3fa-525400199b09:26' at master log mysql-bin.000012, end_log_pos 980. See error log and/or performance_schema.replication_applier_status_by_worker table for more details about this failure or others, if any.

Skip_Counter: 0

分析：

重啟主庫，並且訪問a.b表時， mysqlbinlog中會記錄下面的內容：

# at 327

#170518 19:16:07 server id 10 end_log_pos 410 CRC32 0xc908da54 Query thread_id=5 exec_time=499 error_code=0

SET TIMESTAMP=1495106167/*!*/;

DELETE FROM `a`.`b`

/*!*/;

通過跟蹤可知：

#0 open_table_entry_fini (thd=0x7f3f1c000d80, share=0x7f3f14045b50, entry=0x7f3f1c00eef0) at /data/mysql-5.7.17/sql/sql_base.cc:4325

#1 0x00000000014968ac in open_table (thd=0x7f3f1c000d80, table_list=0x7f3f1c006580, ot_ctx=0x7f3f40092370) at /data/mysql-5.7.17/sql/sql_base.cc:3551

#2 0x000000000149912c in open_and_process_table (thd=0x7f3f1c000d80, lex=0x7f3f1c003078, tables=0x7f3f1c006580, counter=0x7f3f1c003138, flags=0, prelocking_strategy=0x7f3f40092470,

has_prelocking_list=false, ot_ctx=0x7f3f40092370) at /data/mysql-5.7.17/sql/sql_base.cc:5108

#3 0x000000000149a1ce in open_tables (thd=0x7f3f1c000d80, start=0x7f3f40092430, counter=0x7f3f1c003138, flags=0, prelocking_strategy=0x7f3f40092470) at /data/mysql-5.7.17/sql/sql_base.cc:5719

#4 0x000000000149b4fb in open_tables_for_query (thd=0x7f3f1c000d80, tables=0x7f3f1c006580, flags=0) at /data/mysql-5.7.17/sql/sql_base.cc:6494

#5 0x00000000015208fe in execute_sqlcom_select (thd=0x7f3f1c000d80, all_tables=0x7f3f1c006580) at /data/mysql-5.7.17/sql/sql_parse.cc:5166

#6 0x000000000151a193 in mysql_execute_command (thd=0x7f3f1c000d80, first_level=true) at /data/mysql-5.7.17/sql/sql_parse.cc:2794

#7 0x00000000015218f6 in mysql_parse (thd=0x7f3f1c000d80, parser_state=0x7f3f40093690) at /data/mysql-5.7.17/sql/sql_parse.cc:5611

#8 0x000000000151709c in dispatch_command (thd=0x7f3f1c000d80, com_data=0x7f3f40093df0, command=COM_QUERY) at /data/mysql-5.7.17/sql/sql_parse.cc:1461

#9 0x0000000001515f8e in do_command (thd=0x7f3f1c000d80) at /data/mysql-5.7.17/sql/sql_parse.cc:999

#10 0x0000000001645460 in handle_connection (arg=0x43de0e0) at /data/mysql-5.7.17/sql/conn_handler/connection_handler_per_thread.cc:300

#11 0x0000000001cbe494 in pfs_spawn_thread (arg=0x4474250) at /data/mysql-5.7.17/storage/perfschema/pfs.cc:2188

#12 0x00007f3f4e629dc5 in start_thread () from /lib64/libpthread.so.0

#13 0x00007f3f4d4f473d in clone () from /lib64/libc.so.6

是在sql/sql_base.cc:open_table_entry_fini函式中實現的新增delete

具體是體現在：

4345 if (mysql_bin_log.is_open())

4346 {

4347 bool error= false;

4348 String temp_buf;

4349 error= temp_buf.append("DELETE FROM ");

4350 append_identifier(thd, &temp_buf, share->db.str, strlen(share->db.str));

4351 error= temp_buf.append(".");

4352 append_identifier(thd, &temp_buf, share->table_name.str,

4353 strlen(share->table_name.str));

4354 if (mysql_bin_log.write_dml_directly(thd, temp_buf.c_ptr_safe(),

4355 temp_buf.length()))

4356 return TRUE;

4357 if (error)

4358 {

4359 /*

4360 As replication is maybe going to be corrupted, we need to warn the

4361 DBA on top of warning the client (which will automatically be done

4362 because of MYF(MY_WME) in my_malloc() above).

4363 */

4364 sql_print_error("When opening HEAP table, could not allocate memory "

4365 "to write 'DELETE FROM `%s`.`%s`' to the binary log",

4366 share->db.str, share->table_name.str);

4367 delete entry->triggers;

4368 return TRUE;

4369 }

直接在mysqlbinlog中寫了delete 語句。

對於有memory引擎的表，做主從時可能會有的問題：

1 邏輯備份：會dump出來當時memory表裡有的內容，但他不支援事務，所以有可能資料是不一樣的

2 物理備份：memory的表只有frm檔案，所以裡面的內容一定是空的

所以兩種方式都不能保證可以直接做出來從庫。

最好的方法是：在主從複製中還是不要使用memory引擎的表。真要用的話，需要了解他的影響。

宣告：原始碼新手，僅代表個人意見，非權威，如果誤導看官，見諒。

QQ 273002188 歡迎一起學習

QQ 群 236941212

oracle,mysql,PG 相互交流

來自 “ ITPUB部落格 ” ，連結：http://blog.itpub.net/25099483/viewspace-2139311/，如需轉載，請註明出處，否則將追究法律責任。

mysql複製和記憶體引擎的表

相關文章