如何在線修改千萬,億級別的mysql數據庫表的結構方案

在開發時，常常遇到開始在設計數據庫時沒有評估好mysql的數據結構，好比須要臨時新增，修改等對mysql作一些DDL操做，如何數據量百萬級別如下，直接操做mysql命令來修改表結構，問題不大，但當數據庫已經上百萬，千萬，甚至上億了，也沒有分庫分表，須要對這麼大記錄集的表進行修改，若是直接用mysql命令操做，勢必會鎖表時間太長，影響線上業務及時mysql操做使用，以下介紹一種方案解決在線修改大數據表而不影響在線業務的方法：

環境版本： ubuntu 12, mysql 5.5

安裝percona-toolkit工具：

gpg --keyserver  hkp://keys.gnupg.net --recv-keys 1C4CBDCDCD2EFD2A
gpg -a --export CD2EFD2A | sudo apt-key add -
echo "deb http://repo.percona.com/apt `lsb_release -cs` main" >> /etc/apt/sources.list.d/percona.list
 echo "deb-src http://repo.percona.com/apt `lsb_release -cs` main" >> /etc/apt/sources.list.d/percona.list

apt-get update
apt-get install percona-toolkit

安裝percona-toolkit成功後，執行pt-online-schema-change命令，好比須要修改mysql表的字段類型：

./pt-online-schema-change --user=root -h 192.168.0.2  -p root --alter="MODIFY COLUMN count  int(11) UNSIGNED NOT NULL AFTER content"  D=pt-demo,t=news --execute

若是能收到以下輸出，說明執行成功了

No slaves found.  See --recursion-method if host sdk-dev-dbserver has slaves.
Not checking slave lag because no slaves were found and --check-slave-lag was not specified.
Operation, tries, wait:
  analyze_table, 10, 1
  copy_rows, 10, 0.25
  create_triggers, 10, 1
  drop_triggers, 10, 1
  swap_tables, 10, 1
  update_foreign_keys, 10, 1
Altering `pt-demo`.`news`...
Creating new table...
Created new table pt-demo._news_new OK.
Altering new table...
Altered `pt-demo`.`_news_new` OK.
2017-02-09T11:30:18 Creating triggers...
2017-02-09T11:30:18 Created triggers OK.
2017-02-09T11:30:18 Copying approximately 27906818 rows...
Copying `pt-demo`.`news`:   4% 09:51 remain
Copying `pt-demo`.`news`:   9% 09:40 remain
Copying `pt-demo`.`news`:  13% 09:36 remain
Copying `pt-demo`.`news`:  17% 09:15 remain
Copying `pt-demo`.`news`:  22% 08:50 remain
Copying `pt-demo`.`news`:  26% 08:16 remain
Copying `pt-demo`.`news`:  33% 06:50 remain
Copying `pt-demo`.`news`:  39% 06:05 remain
Copying `pt-demo`.`news`:  43% 05:45 remain
Copying `pt-demo`.`news`:  48% 05:23 remain
Copying `pt-demo`.`news`:  51% 05:07 remain
Copying `pt-demo`.`news`:  55% 04:44 remain
Copying `pt-demo`.`news`:  60% 04:18 remain
Copying `pt-demo`.`news`:  66% 03:34 remain
Copying `pt-demo`.`news`:  72% 02:49 remain
Copying `pt-demo`.`news`:  79% 02:02 remain
Copying `pt-demo`.`news`:  86% 01:19 remain
Copying `pt-demo`.`news`:  93% 00:37 remain
Copying `pt-demo`.`news`:  98% 00:07 remain
2017-02-09T11:40:54 Copied rows OK.
2017-02-09T11:40:54 Swapping tables...
2017-02-09T11:40:54 Swapped original and new tables OK.
2017-02-09T11:40:54 Dropping old table...
2017-02-09T11:40:55 Dropped old table `pt-demo`.`_news_old` OK.
2017-02-09T11:40:55 Dropping triggers...
2017-02-09T11:40:55 Dropped triggers OK.
Successfully altered `pt-demo`.`news`.

 

percona-toolkit工做原理是，1，先建立需修改字段後的臨時表；2，把舊數據導入到臨時表；3，建立了觸發器，把在執行修改結構時對舊的表數據操做觸發導入到臨時表；4，把臨時表命名成舊錶名稱。