MySQL 中加速 LEFT JOIN 的建议
Suggestions for speeding up LEFT JOIN in MySQL
我有一个名为 usr_data 的 tabled，其中有 40,000 条左右的记录，我想通过 select 在加入，显然这真的很慢（最多150秒）。
我想知道是否有任何方法可以加快速度？不幸的是，这是连接中子 select 的最快查询。
我的查询
SELECT usr_data.usr_id,usr_data.login,orgus.title FROM usr_data
LEFT JOIN (
SELECT object_reference.ref_id,rbac_ua.usr_id,object_data.obj_id,object_data.title
FROM rbac_ua
JOIN rbac_fa ON rbac_fa.rol_id = rbac_ua.rol_id
JOIN object_reference ON rbac_fa.parent = object_reference.ref_id
JOIN object_data ON object_data.obj_id = object_reference.obj_id
JOIN object_data role ON role.obj_id = rbac_ua.rol_id
WHERE object_data.type = 'orgu') as orgus on orgus.usr_id = usr_data.usr_id
WHERE usr_data.usr_id > 0 AND usr_data.login <> "anonymous"
查询描述
+------+-------------+------------------+--------+-----------------------+---------+---------+--------------------------------+------+--------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+------+-------------+------------------+--------+-----------------------+---------+---------+--------------------------------+------+--------------------------+
| 1 | SIMPLE | usr_data | range | PRIMARY,i1_idx | PRIMARY | 4 | NULL | 8148 | Using where |
| 1 | SIMPLE | rbac_ua | ref | PRIMARY,i1_idx,i2_idx | PRIMARY | 4 | ildCPC.usr_data.usr_id | 2 | Using where; Using index |
| 1 | SIMPLE | rbac_fa | ref | PRIMARY,i1_idx | PRIMARY | 4 | ildCPC.rbac_ua.rol_id | 1 | Using where; Using index |
| 1 | SIMPLE | role | eq_ref | PRIMARY | PRIMARY | 4 | ildCPC.rbac_ua.rol_id | 1 | Using index |
| 1 | SIMPLE | object_reference | eq_ref | PRIMARY,i1_idx | PRIMARY | 4 | ildCPC.rbac_fa.parent | 1 | Using where |
| 1 | SIMPLE | object_data | eq_ref | PRIMARY,i1_idx | PRIMARY | 4 | ildCPC.object_reference.obj_id | 1 | Using where |
+------+-------------+------------------+--------+-----------------------+---------+---------+--------------------------------+------+--------------------------+
usr_data table
+----------------------+---------------+------+-----+---------+-------+
| Field | Type | Null | Key | Default | Extra |
+----------------------+---------------+------+-----+---------+-------+
| usr_id | int(11) | NO | PRI | 0 | |
| login | varchar(80) | YES | MUL | NULL | |
| passwd | varchar(80) | YES | | NULL | |
| firstname | varchar(32) | YES | | NULL | |
| lastname | varchar(32) | YES | | NULL | |
| title | varchar(32) | YES | | NULL | |
| gender | char(1) | YES | | m | |
| email | varchar(80) | YES | | NULL | |
usr_data 索引
+----------+------------+----------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
| Table | Non_unique | Key_name | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type | Comment | Index_comment |
+----------+------------+----------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
| usr_data | 0 | PRIMARY | 1 | usr_id | A | 29354 | NULL | NULL | | BTREE | | |
| usr_data | 1 | i1_idx | 1 | login | A | 29354 | NULL | NULL | YES | BTREE | | |
| usr_data | 1 | i1_idx | 2 | passwd | A | 29354 | NULL | NULL | YES | BTREE | | |
| usr_data | 1 | i2_idx | 1 | ext_account | A | 2 | NULL | NULL | YES | BTREE | | |
| usr_data | 1 | i2_idx | 2 | auth_mode | A | 4 | NULL | NULL | YES | BTREE | | |
+----------+------------+----------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
我尝试在 LEFT JOIN 中为 SELECT 创建一个临时 table，但它并没有真正加快速度，目前查询最多需要 150 秒，通过正确的连接，它可以将其降低到大约 1 秒。 （这是较小的table）。
select * from object_data where type = 'orgu' returns 1058 rows.
显示创建 TABLE rbac_ua
+---------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Table | Create Table |
+---------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| rbac_ua | CREATE TABLE `rbac_ua` (
`usr_id` int(11) NOT NULL DEFAULT '0',
`rol_id` int(11) NOT NULL DEFAULT '0',
PRIMARY KEY (`usr_id`,`rol_id`),
KEY `i1_idx` (`usr_id`),
KEY `i2_idx` (`rol_id`),
KEY `rol_id` (`rol_id`,`usr_id`),
KEY `rol_usr` (`rol_id`,`usr_id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci ROW_FORMAT=DYNAMIC |
+---------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
显示创建 TABLE rbac_fa
+---------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Table | Create Table |
+---------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| rbac_fa | CREATE TABLE `rbac_fa` (
`rol_id` int(11) NOT NULL DEFAULT '0',
`parent` int(11) NOT NULL DEFAULT '0',
`assign` char(1) COLLATE utf8_unicode_ci DEFAULT NULL,
`protected` char(1) COLLATE utf8_unicode_ci DEFAULT 'n',
PRIMARY KEY (`rol_id`,`parent`),
KEY `i1_idx` (`parent`),
KEY `parent` (`parent`,`rol_id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci ROW_FORMAT=DYNAMIC |
+---------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
显示创建 TABLE object_data
+-------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Table | Create Table |
+-------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| object_data | CREATE TABLE `object_data` (
`obj_id` int(11) NOT NULL DEFAULT '0',
`type` char(4) COLLATE utf8_unicode_ci DEFAULT 'none',
`title` char(255) COLLATE utf8_unicode_ci DEFAULT NULL,
`description` char(128) COLLATE utf8_unicode_ci DEFAULT NULL,
`owner` int(11) NOT NULL DEFAULT '0',
`create_date` datetime DEFAULT NULL,
`last_update` datetime DEFAULT NULL,
`import_id` char(50) COLLATE utf8_unicode_ci DEFAULT NULL,
PRIMARY KEY (`obj_id`),
KEY `i1_idx` (`type`),
KEY `i2_idx` (`title`),
KEY `i4_idx` (`import_id`),
FULLTEXT KEY `i3_idx` (`title`,`description`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci ROW_FORMAT=DYNAMIC |
+-------------+--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
显示创建 TABLE object_reference
+------------------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Table | Create Table |
+------------------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| object_reference | CREATE TABLE `object_reference` (
`ref_id` int(11) NOT NULL DEFAULT '0',
`obj_id` int(11) NOT NULL DEFAULT '0',
`deleted` datetime DEFAULT NULL,
PRIMARY KEY (`ref_id`),
KEY `i1_idx` (`obj_id`),
KEY `i2_idx` (`deleted`),
KEY `obj_id` (`obj_id`,`ref_id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci ROW_FORMAT=DYNAMIC |
+------------------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
如果我遗漏了任何有用的信息，请告诉我。
提前致谢。
通过删除子选择和使用连接，您应该能够提高性能。首先处理子选择，生成一个临时文件 table，这是一个相当大的性能损失。
SELECT
usr_data.usr_id,
usr_data.login,
object_data.title,
object_reference.ref_id,
rbac_ua.usr_id,
object_data.obj_id,
object_data.title
FROM usr_data
LEFT JOIN rbac_ua
ON rbac_ua.usr_id = usr_data.usr_id
JOIN rbac_fa ON rbac_fa.rol_id = rbac_ua.rol_id
JOIN object_reference ON rbac_fa.parent = object_reference.ref_id
JOIN object_data ON object_data.obj_id = object_reference.obj_id
JOIN object_data role ON role.obj_id = rbac_ua.rol_id
WHERE
object_data.type = 'orgu' AND
usr_data.usr_id > 0 AND
usr_data.login <> "anonymous"
通过使用 LEFT JOIN，您需要一个大型结果集，其中包含与您的 WHERE usr_data whatever 子句匹配的每个项目的行。即使查询规划器做得很好，这也需要时间推送到您的客户端软件。
usr_data.usr_id > 0 似乎是多余的，因为您的 usr_id 列似乎是一个自动增量列。所有值都大于零吗？
子选择绝对是一个性能消耗者。
先生Thrasher 很接近，我想。问题是在 WHERE 子句中提到 LEFT JOINed table 的列：将 LEFT JOIN 转换为直接的 JOIN.
试试这个看看你是否得到合适的结果。我不明白你的架构，所以这里有一些猜测。
SELECT usr_data.usr_id, usr_data.login, object_data.title
FROM usr_data
LEFT JOIN rbac_ua ON rbac_ua.usr_id = usr_data.usr_id
LEFT JOIN rbac_fa ON rbac_fa.rol_id = rbac_ua.rol_id
LEFT JOIN object_reference ON rbac_fa.parent = object_reference.ref_id
LEFT JOIN object_data ON object_data.obj_id = object_reference.obj_id
AND object_data.type = 'orgu'
LEFT JOIN object_data role ON role.obj_id = rbac_ua.rol_id
WHERE usr_data.login <> 'anonymous'
注意 object_data.type = 'orgu' 如何在 ON 子句中结束。 （是的，ON 子句可以包含与 WHERE 子句相同的内容！）这使得它不会将 LEFT JOIN 变成直接的 JOIN.
我不确定这行的重点。
LEFT JOIN object_data role ON role.obj_id = rbac_ua.rol_id
table 似乎对您的结果集没有贡献。
从 EXPLAIN 中，我看到优化程序认为“LEFT”没有影响。所以它删除了它。然后它决定 "derived table" 可以变成 JOIN。你本可以做到这一切的。
但真正的性能问题是过滤的重要部分在 type = 'orgu' 上，但被埋在 'last' table.
中
（此后我猜测是因为 每个 table 都缺少 SHOW CREATE TABLE。而 DESCRIBE 不像描述性广告 SHOW CREATE.)
让我们看看这个：
SELECT ud.usr_id, ud.login, od.title
FROM object_data AS od
JOIN object_reference AS r ON od.obj_id = r.obj_id
JOIN rbac_fa AS rfa ON rfa.parent = r.ref_id
JOIN rbac_ua AS rua ON rfa.rol_id = rua.rol_id
JOIN usr_data AS ud ON ud.usr_id = rua.usr_id
-- unnec?: JOIN object_data role ON role.obj_id = rua.rol_id
WHERE od.type = 'orgu'
AND ud.usr_id > 0
AND ud.login <> "anonymous"
我认为在 添加下面的索引后效率更高。如果 type = 'orgu' 具有足够的选择性，则可能会出现这种情况。
请注意，role 除了验证给定的 obj_id.
是否存在角色外，不向查询添加任何内容
需要索引以便它可以 start with type = 'orgu':
object_data: INDEX(type)
object_reference: INDEX(obj_id, ref_id) -- (covering, too)
rbac_fa: INDEX(parent, rol_id) -- (covering, too)
rbac_ua: INDEX(rol_id, usr_id) -- (covering, too)
如果没有帮助，请提供 EXPLAIN SELECT ... 结果。并告诉我们有多少行 object_data 有 type=orgu.
我有一个名为 usr_data 的 tabled，其中有 40,000 条左右的记录，我想通过 select 在加入，显然这真的很慢（最多150秒）。
我想知道是否有任何方法可以加快速度？不幸的是，这是连接中子 select 的最快查询。
我的查询
SELECT usr_data.usr_id,usr_data.login,orgus.title FROM usr_data
LEFT JOIN (
SELECT object_reference.ref_id,rbac_ua.usr_id,object_data.obj_id,object_data.title
FROM rbac_ua
JOIN rbac_fa ON rbac_fa.rol_id = rbac_ua.rol_id
JOIN object_reference ON rbac_fa.parent = object_reference.ref_id
JOIN object_data ON object_data.obj_id = object_reference.obj_id
JOIN object_data role ON role.obj_id = rbac_ua.rol_id
WHERE object_data.type = 'orgu') as orgus on orgus.usr_id = usr_data.usr_id
WHERE usr_data.usr_id > 0 AND usr_data.login <> "anonymous"
查询描述
+------+-------------+------------------+--------+-----------------------+---------+---------+--------------------------------+------+--------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+------+-------------+------------------+--------+-----------------------+---------+---------+--------------------------------+------+--------------------------+
| 1 | SIMPLE | usr_data | range | PRIMARY,i1_idx | PRIMARY | 4 | NULL | 8148 | Using where |
| 1 | SIMPLE | rbac_ua | ref | PRIMARY,i1_idx,i2_idx | PRIMARY | 4 | ildCPC.usr_data.usr_id | 2 | Using where; Using index |
| 1 | SIMPLE | rbac_fa | ref | PRIMARY,i1_idx | PRIMARY | 4 | ildCPC.rbac_ua.rol_id | 1 | Using where; Using index |
| 1 | SIMPLE | role | eq_ref | PRIMARY | PRIMARY | 4 | ildCPC.rbac_ua.rol_id | 1 | Using index |
| 1 | SIMPLE | object_reference | eq_ref | PRIMARY,i1_idx | PRIMARY | 4 | ildCPC.rbac_fa.parent | 1 | Using where |
| 1 | SIMPLE | object_data | eq_ref | PRIMARY,i1_idx | PRIMARY | 4 | ildCPC.object_reference.obj_id | 1 | Using where |
+------+-------------+------------------+--------+-----------------------+---------+---------+--------------------------------+------+--------------------------+
usr_data table
+----------------------+---------------+------+-----+---------+-------+
| Field | Type | Null | Key | Default | Extra |
+----------------------+---------------+------+-----+---------+-------+
| usr_id | int(11) | NO | PRI | 0 | |
| login | varchar(80) | YES | MUL | NULL | |
| passwd | varchar(80) | YES | | NULL | |
| firstname | varchar(32) | YES | | NULL | |
| lastname | varchar(32) | YES | | NULL | |
| title | varchar(32) | YES | | NULL | |
| gender | char(1) | YES | | m | |
| email | varchar(80) | YES | | NULL | |
usr_data 索引
+----------+------------+----------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
| Table | Non_unique | Key_name | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type | Comment | Index_comment |
+----------+------------+----------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
| usr_data | 0 | PRIMARY | 1 | usr_id | A | 29354 | NULL | NULL | | BTREE | | |
| usr_data | 1 | i1_idx | 1 | login | A | 29354 | NULL | NULL | YES | BTREE | | |
| usr_data | 1 | i1_idx | 2 | passwd | A | 29354 | NULL | NULL | YES | BTREE | | |
| usr_data | 1 | i2_idx | 1 | ext_account | A | 2 | NULL | NULL | YES | BTREE | | |
| usr_data | 1 | i2_idx | 2 | auth_mode | A | 4 | NULL | NULL | YES | BTREE | | |
+----------+------------+----------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
我尝试在 LEFT JOIN 中为 SELECT 创建一个临时 table，但它并没有真正加快速度，目前查询最多需要 150 秒，通过正确的连接，它可以将其降低到大约 1 秒。 （这是较小的table）。
select * from object_data where type = 'orgu' returns 1058 rows.
显示创建 TABLE rbac_ua
+---------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Table | Create Table |
+---------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| rbac_ua | CREATE TABLE `rbac_ua` (
`usr_id` int(11) NOT NULL DEFAULT '0',
`rol_id` int(11) NOT NULL DEFAULT '0',
PRIMARY KEY (`usr_id`,`rol_id`),
KEY `i1_idx` (`usr_id`),
KEY `i2_idx` (`rol_id`),
KEY `rol_id` (`rol_id`,`usr_id`),
KEY `rol_usr` (`rol_id`,`usr_id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci ROW_FORMAT=DYNAMIC |
+---------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
显示创建 TABLE rbac_fa
+---------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Table | Create Table |
+---------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| rbac_fa | CREATE TABLE `rbac_fa` (
`rol_id` int(11) NOT NULL DEFAULT '0',
`parent` int(11) NOT NULL DEFAULT '0',
`assign` char(1) COLLATE utf8_unicode_ci DEFAULT NULL,
`protected` char(1) COLLATE utf8_unicode_ci DEFAULT 'n',
PRIMARY KEY (`rol_id`,`parent`),
KEY `i1_idx` (`parent`),
KEY `parent` (`parent`,`rol_id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci ROW_FORMAT=DYNAMIC |
+---------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
显示创建 TABLE object_data
+-------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Table | Create Table |
+-------------+----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| object_data | CREATE TABLE `object_data` (
`obj_id` int(11) NOT NULL DEFAULT '0',
`type` char(4) COLLATE utf8_unicode_ci DEFAULT 'none',
`title` char(255) COLLATE utf8_unicode_ci DEFAULT NULL,
`description` char(128) COLLATE utf8_unicode_ci DEFAULT NULL,
`owner` int(11) NOT NULL DEFAULT '0',
`create_date` datetime DEFAULT NULL,
`last_update` datetime DEFAULT NULL,
`import_id` char(50) COLLATE utf8_unicode_ci DEFAULT NULL,
PRIMARY KEY (`obj_id`),
KEY `i1_idx` (`type`),
KEY `i2_idx` (`title`),
KEY `i4_idx` (`import_id`),
FULLTEXT KEY `i3_idx` (`title`,`description`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci ROW_FORMAT=DYNAMIC |
+-------------+--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
显示创建 TABLE object_reference
+------------------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Table | Create Table |
+------------------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| object_reference | CREATE TABLE `object_reference` (
`ref_id` int(11) NOT NULL DEFAULT '0',
`obj_id` int(11) NOT NULL DEFAULT '0',
`deleted` datetime DEFAULT NULL,
PRIMARY KEY (`ref_id`),
KEY `i1_idx` (`obj_id`),
KEY `i2_idx` (`deleted`),
KEY `obj_id` (`obj_id`,`ref_id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci ROW_FORMAT=DYNAMIC |
+------------------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
如果我遗漏了任何有用的信息，请告诉我。
提前致谢。
通过删除子选择和使用连接，您应该能够提高性能。首先处理子选择，生成一个临时文件 table，这是一个相当大的性能损失。
SELECT
usr_data.usr_id,
usr_data.login,
object_data.title,
object_reference.ref_id,
rbac_ua.usr_id,
object_data.obj_id,
object_data.title
FROM usr_data
LEFT JOIN rbac_ua
ON rbac_ua.usr_id = usr_data.usr_id
JOIN rbac_fa ON rbac_fa.rol_id = rbac_ua.rol_id
JOIN object_reference ON rbac_fa.parent = object_reference.ref_id
JOIN object_data ON object_data.obj_id = object_reference.obj_id
JOIN object_data role ON role.obj_id = rbac_ua.rol_id
WHERE
object_data.type = 'orgu' AND
usr_data.usr_id > 0 AND
usr_data.login <> "anonymous"
通过使用 LEFT JOIN，您需要一个大型结果集，其中包含与您的 WHERE usr_data whatever 子句匹配的每个项目的行。即使查询规划器做得很好，这也需要时间推送到您的客户端软件。
usr_data.usr_id > 0 似乎是多余的，因为您的 usr_id 列似乎是一个自动增量列。所有值都大于零吗？
子选择绝对是一个性能消耗者。
先生Thrasher 很接近，我想。问题是在 WHERE 子句中提到 LEFT JOINed table 的列：将 LEFT JOIN 转换为直接的 JOIN.
试试这个看看你是否得到合适的结果。我不明白你的架构，所以这里有一些猜测。
SELECT usr_data.usr_id, usr_data.login, object_data.title
FROM usr_data
LEFT JOIN rbac_ua ON rbac_ua.usr_id = usr_data.usr_id
LEFT JOIN rbac_fa ON rbac_fa.rol_id = rbac_ua.rol_id
LEFT JOIN object_reference ON rbac_fa.parent = object_reference.ref_id
LEFT JOIN object_data ON object_data.obj_id = object_reference.obj_id
AND object_data.type = 'orgu'
LEFT JOIN object_data role ON role.obj_id = rbac_ua.rol_id
WHERE usr_data.login <> 'anonymous'
注意 object_data.type = 'orgu' 如何在 ON 子句中结束。 （是的，ON 子句可以包含与 WHERE 子句相同的内容！）这使得它不会将 LEFT JOIN 变成直接的 JOIN.
我不确定这行的重点。
LEFT JOIN object_data role ON role.obj_id = rbac_ua.rol_id
table 似乎对您的结果集没有贡献。
从 EXPLAIN 中，我看到优化程序认为“LEFT”没有影响。所以它删除了它。然后它决定 "derived table" 可以变成 JOIN。你本可以做到这一切的。
但真正的性能问题是过滤的重要部分在 type = 'orgu' 上，但被埋在 'last' table.
（此后我猜测是因为 每个 table 都缺少 SHOW CREATE TABLE。而 DESCRIBE 不像描述性广告 SHOW CREATE.)
让我们看看这个：
SELECT ud.usr_id, ud.login, od.title
FROM object_data AS od
JOIN object_reference AS r ON od.obj_id = r.obj_id
JOIN rbac_fa AS rfa ON rfa.parent = r.ref_id
JOIN rbac_ua AS rua ON rfa.rol_id = rua.rol_id
JOIN usr_data AS ud ON ud.usr_id = rua.usr_id
-- unnec?: JOIN object_data role ON role.obj_id = rua.rol_id
WHERE od.type = 'orgu'
AND ud.usr_id > 0
AND ud.login <> "anonymous"
我认为在 添加下面的索引后效率更高。如果 type = 'orgu' 具有足够的选择性，则可能会出现这种情况。
请注意，role 除了验证给定的 obj_id.
需要索引以便它可以 start with type = 'orgu':
object_data: INDEX(type)
object_reference: INDEX(obj_id, ref_id) -- (covering, too)
rbac_fa: INDEX(parent, rol_id) -- (covering, too)
rbac_ua: INDEX(rol_id, usr_id) -- (covering, too)
如果没有帮助，请提供 EXPLAIN SELECT ... 结果。并告诉我们有多少行 object_data 有 type=orgu.