1. 技术文章
  2. Slow auto_increment reset

Slow auto_increment reset

由 隐身守侯 提交于 2019-12-03 15:59:31

simply add a temporary record that has desired auto_increment_id-1 to each table, and remove the record after that, quick and easy, but probably too dirty

example:

insert into item set id=1097972232-1;

after the execution, the next auto_increment will be 1097972232, which is what you desired

this can avoid slowness

This is a documented "feature" of MySQL:

If you use any option to ALTER TABLE other than RENAME, MySQL always creates a temporary table, even if the data wouldn't strictly need to be copied (such as when you change the name of a column). For MyISAM tables, you can speed up the index re-creation operation (which is the slowest part of the alteration process) by setting the myisam_sort_buffer_size system variable to a high value.

http://dev.mysql.com/doc/refman/5.0/en/alter-table.html

MySQL 5.1 and 5.5 support a few more alter table operations w/o a temporary table, but changing the auto_increment is not documented to be one of those.

Why do you need to change the auto_increment value, anyway? This isn't something you should be doing routinely.

There is no simple way to get around the AUTO_INCREMENT attribute default behavior in MySQL, and even if you find a way, I wouldn't recommend you to do so, since it is the best way to run into problems in the short-term. AUTO_INCREMENT values are not meant to be adjusted or reseted in a production environment.

One possible solution to your problem could be to denormalize your model a little. The idea is to move the AUTO_INCREMENT field to a side table where you don't have to copy or delete rows. All you have to do then is to get a new id value from this side table when you create a new item, and keep the existing id value when you copy rows from one table to another.

To achieve this we will use a trigger that will create a new id for us and assign it to our item record. The id field of the item table has to be nullable for this to work, so we have to replace the primary key by a unique index.

This model change would be completely transparent for your application, so you would have no change to make in your application code.

Here are some example scripts. Let's say we have two item tables in our database, with some common rows, and some rows that need to be moved from first table to second table:

CREATE TABLE `item1` (
  `id` bigint(20) unsigned NOT NULL AUTO_INCREMENT,
  `item_res_id` int(11) NOT NULL DEFAULT '0',
  PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;

CREATE TABLE `item2` (
  `id` bigint(20) unsigned NOT NULL AUTO_INCREMENT,
  `item_res_id` int(11) NOT NULL DEFAULT '0',
  PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;

INSERT INTO item1 (item_res_id) VALUES (1);
INSERT INTO item1 (item_res_id) VALUES (2);
INSERT INTO item2 (item_res_id) VALUES (1);

If we try to move some data from one table to another and then restart your server, we will encounter the issue of AUTO_INCREMENT value reset. So we will slightly modify our model as follows:

We will proceed in several steps to migrate our data model. The DDL statements in the following migration scripts have been generated using neXtep Designer IDE.

  • First we create a new item_keys table that will hold the AUTO_INCREMENT field:
-- Creating table 'item_keys'
CREATE TABLE item_keys ( 
   id BIGINT(20) UNSIGNED NOT NULL
  ,key_ctime TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
) Engine=InnoDB default charset=utf8;

-- Creating Primary Key constraint 'PRIMARY' on table 'item_keys'
ALTER TABLE item_keys ADD CONSTRAINT PRIMARY KEY (id);
  • But before activating the AUTO_INCREMENT attribute, we must insert existing ids into our new table:
-- Initializing item_keys with existing ids
INSERT INTO item_keys (id)
    SELECT i1.id
    FROM item1 i1
        LEFT JOIN item_keys ik ON ik.id = i1.id
    WHERE ik.id IS NULL
;

INSERT INTO item_keys (id)
    SELECT i2.id
    FROM item2 i2
        LEFT JOIN item_keys ik ON ik.id = i2.id
    WHERE ik.id IS NULL
;
  • We can now activate the AUTO_INCREMENT attribute, and initialize its value for future inserts:
-- Activating auto_increment constraint...
ALTER TABLE item_keys MODIFY id BIGINT(20) UNSIGNED NOT NULL AUTO_INCREMENT;

-- Initializing auto_increment value
SELECT @inc_value := MAX(id) FROM item_keys;
SET @alter_query = CONCAT('ALTER TABLE item_keys AUTO_INCREMENT=',@inc_value); 
PREPARE alter_query FROM @alter_query; 
EXECUTE alter_query; 
DEALLOCATE PREPARE alter_query;
  • Then we can alter the item1 and item2 tables to replace the primary key by a unique index, and reference the primary key of the item_keys table:
-- De-activating auto_increment constraint...
ALTER TABLE item1 MODIFY id BIGINT(20) UNSIGNED NOT NULL;
-- Dropping constraint 'PRIMARY'...
ALTER TABLE item1 DROP PRIMARY KEY;
ALTER TABLE item1 MODIFY id BIGINT(20) UNSIGNED NULL;
-- Creating index 'item1_uk'...
CREATE UNIQUE INDEX item1_uk ON item1 (id);
-- Creating Foreign Key constraint 'item1_keys_fk' on table 'item1'
ALTER TABLE item1 ADD 
   CONSTRAINT item1_keys_fk FOREIGN KEY item1_keys_fk
      (id) REFERENCES item_keys
      (id)
;

-- De-activating auto_increment constraint...
ALTER TABLE item2 MODIFY id BIGINT(20) UNSIGNED NOT NULL;
-- Dropping constraint 'PRIMARY'...
ALTER TABLE item2 DROP PRIMARY KEY;
ALTER TABLE item2 MODIFY id BIGINT(20) UNSIGNED NULL;
-- Creating index 'item2_uk'...
CREATE UNIQUE INDEX item2_uk ON item2 (id);
-- Creating Foreign Key constraint 'item2_keys_fk' on table 'item2'
ALTER TABLE item2 ADD 
   CONSTRAINT item2_keys_fk FOREIGN KEY item2_keys_fk
      (id) REFERENCES item_keys
      (id)
;
  • Finally, we just have to create the triggers that will manage the ids creation for us:
-- Creating trigger 'tr_item1_bi' on table 'item1'...
DELIMITER |;
CREATE TRIGGER tr_item1_bi BEFORE INSERT ON item1
FOR EACH ROW
BEGIN
   IF (NEW.id IS NULL) THEN

        -- If no item id has been specified in the INSERT statement, it
        -- means we want to create a new item. We insert a new record
        -- into the item_keys table to get an item id.
        INSERT INTO item_keys (
            key_ctime
          )
        VALUES (NOW());

        SET NEW.id = LAST_INSERT_ID();
    END IF;
END;
|;

-- Creating trigger 'tr_item2_bi' on table 'item2'...
DELIMITER |;
CREATE TRIGGER tr_item2_bi BEFORE INSERT ON item2
FOR EACH ROW
BEGIN
   IF (NEW.id IS NULL) THEN

        -- If no item id has been specified in the INSERT statement, it
        -- means we want to create a new item. We insert a new record
        -- into the item_keys table to get an item id.
        INSERT INTO item_keys (
            key_ctime
          )
        VALUES (NOW());

        SET NEW.id = LAST_INSERT_ID();
    END IF;
END;
|;

Now we can move data from one table to another, keeping the ids unchanged, and if we restart the server, the AUTO_INCREMENT value in the item_keys will stay the same.

--------------
INSERT INTO item2
    SELECT i1.*
    FROM item1 i1
        LEFT JOIN item2 i2
            ON i2.id = i1.id
    WHERE i2.id IS NULL
--------------
Query OK, 1 row affected (0.04 sec)
Records: 1  Duplicates: 0  Warnings: 0

--------------
DELETE FROM item1
--------------
Query OK, 2 rows affected (0.00 sec)

--------------
INSERT INTO item1 (item_res_id) VALUES (3)
--------------
Query OK, 1 row affected (0.00 sec)

--------------
SELECT * FROM item1
--------------

+------+-------------+
| id   | item_res_id |
+------+-------------+
|    3 |           3 |
+------+-------------+
1 row in set (0.00 sec)

--------------
SELECT * FROM item2
--------------

+------+-------------+
| id   | item_res_id |
+------+-------------+
|    1 |           1 |
|    2 |           2 |
+------+-------------+
2 rows in set (0.00 sec)

--------------
SELECT * FROM item_keys
--------------

+----+---------------------+
| id | key_ctime           |
+----+---------------------+
|  1 | 2010-11-14 10:31:21 |
|  2 | 2010-11-14 10:31:21 |
|  3 | 2010-11-14 10:31:46 |
+----+---------------------+
3 rows in set (0.00 sec)

If you need to maintain unique IDs between two or more servers, don't use this alter table method of resetting the auto_increment each time. It'd be easier to change the increment's increment so that each server will generate unique IDs without intervention. For two servers, you set one to start at 0 and one to start at 1, with an increment of 2 - afterward one will generate even IDs, the other will generate odds. With 3 or more servers, you just set the initial values to 0/1/2 with increments of 3, for four it's 0/1/2/3 with inc of 4, etc...

Details on the server-side settings here:

http://dev.mysql.com/doc/refman/5.1/en/replication-options-master.html#sysvar_auto_increment_increment

This way you only have to reset the auto_increment once per table per server, and afterward they'll take care of the uniqueness problem automatically.

Isn't it :

ALTER TABLE item AUTO_INCREMENT=1;

?

Source

标签
mysql
InnoDB
auto-increment
易学教程内所有资源均来自网络或用户发布的内容，如有违反法律规定的内容欢迎反馈！
该文章没有解决你所遇到的问题?点击提问,说说你的问题,让更多的人一起探讨吧!