标签:inno   const   处事   大量   its   tab   pen   之间   比较   
http://blog.csdn.net/xj626852095/article/details/52767963
step 1
 使用explain 查看执行计划, 5.6后可以加参数 explain format=json xxx 输出json格式的信息
 
step 2 
使用profiling详细的列出在每一个步骤消耗的时间,前提是先执行一遍语句。
 
#打开profiling 的设置
SET profiling = 1;
SHOW VARIABLES LIKE ‘%profiling%‘;
#查看队列的内容
show profiles;  
#来查看统计信息
show profile block io,cpu for query 3;
step 3  
 
Optimizer trace是MySQL5.6添加的新功能,可以看到大量的内部查询计划产生的信息, 先打开设置,然后执行一次sql,最后查看`information_schema`.`OPTIMIZER_TRACE`的内容
 
#打开设置
SET optimizer_trace=‘enabled=on‘;  
#最大内存根据实际情况而定, 可以不设置
SET OPTIMIZER_TRACE_MAX_MEM_SIZE=1000000;
SET END_MARKERS_IN_JSON=ON;
SET optimizer_trace_limit = 1;
SHOW VARIABLES LIKE ‘%optimizer_trace%‘;
#执行所需sql后,查看该表信息即可看到详细的执行过程
SELECT * FROM `information_schema`.`OPTIMIZER_TRACE`;
 
 
 
MySQL索引选择不正确并详细解析OPTIMIZER_TRACE格式
http://blog.csdn.net/melody_mr/article/details/48950601
一 表结构如下: 
CREATE TABLE t_audit_operate_log (
  Fid bigint(16) AUTO_INCREMENT,
  Fcreate_time int(10) unsigned NOT NULL DEFAULT ‘0‘,
  Fuser varchar(50) DEFAULT ‘‘,
  Fip bigint(16) DEFAULT NULL,
  Foperate_object_id bigint(20) DEFAULT ‘0‘,
  PRIMARY KEY (Fid),
  KEY indx_ctime (Fcreate_time),
  KEY indx_user (Fuser),
  KEY indx_objid (Foperate_object_id),
  KEY indx_ip (Fip)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
执行查询:
MySQL> explain select count(*) from t_audit_operate_log where Fuser=‘XX@XX.com‘ and Fcreate_time>=1407081600 and Fcreate_time<=1407427199\G
*************************** 1. row ***************************
id: 1
select_type: SIMPLE
table: t_audit_operate_log
type: ref
possible_keys: indx_ctime,indx_user
key: indx_user
key_len: 153
ref: const
rows: 2007326
Extra: Using where
 
发现,使用了一个不合适的索引, 不是很理想,于是改成指定索引:
mysql> explain select count(*) from t_audit_operate_log use index(indx_ctime) where Fuser=‘CY6016@cyou-inc.com‘ and Fcreate_time>=1407081600 and Fcreate_time<=1407427199\G
*************************** 1. row ***************************
id: 1
select_type: SIMPLE
table: t_audit_operate_log
type: range
possible_keys: indx_ctime
key: indx_ctime
key_len: 5
ref: NULL
rows: 670092
Extra: Using where
实际执行耗时,后者比前者快了接近10
问题: 很奇怪,优化器为何不选择使用 indx_ctime 索引,而选择了明显会扫描更多行的 indx_user 索引。
分析2个索引的数据量如下:  两个条件的唯一性对比:
select count(*) from t_audit_operate_log where Fuser=‘XX@XX.com‘;
+----------+
| count(*) |
+----------+
| 1238382 | 
+----------+
select count(*) from t_audit_operate_log where Fcreate_time>=1407254400 and Fcreate_time<=1407427199;
+----------+
| count(*) |
+----------+
| 198920 | 
+----------+
显然,使用索引indx_ctime好于indx_user,但MySQL却选择了indx_user. 为什么?
于是,使用 OPTIMIZER_TRACE进一步探索.
 
二  OPTIMIZER_TRACE的过程说明
以本处事例简要说明OPTIMIZER_TRACE的过程.
查看OPTIMIZER_TRACE方法:
1.set optimizer_trace=‘enabled=on‘;    --- 开启trace
2.set optimizer_trace_max_mem_size=1000000;    --- 设置trace大小
3.set end_markers_in_json=on;    --- 增加trace中注释
4.select * from information_schema.optimizer_trace\G;
 
- {\  
-   "steps": [\  
-     {\  
-       "join_preparation": {\  ---优化准备工作  
-         "select#": 1,\  
-         "steps": [\  
-           {\  
-             "expanded_query": "/* select#1 */ select count(0) AS `count(*)` from `t_audit_operate_log` where ((`t_audit_operate_log`.`Fuser` = ‘XX@XX.com‘) and (`t_audit_operate_log`.`Fcreate_time` >= 1407081600) and (`t_audit_operate_log`.`Fcreate_time` <= 1407427199))"\  
-           }\  
-         ] /* steps */\  
-       } /* join_preparation */\  
-     },\  
-     {\  
-       "join_optimization": {\  ---优化工作的主要阶段,包括逻辑优化和物理优化两个阶段  
-         "select#": 1,\  
-         "steps": [\  ---优化工作的主要阶段, 逻辑优化阶段  
-           {\  
-             "condition_processing": {\  ---逻辑优化,条件化简  
-               "condition": "WHERE",\  
-               "original_condition": "((`t_audit_operate_log`.`Fuser` = ‘XX@XX.com‘) and (`t_audit_operate_log`.`Fcreate_time` >= 1407081600) and (`t_audit_operate_log`.`Fcreate_time` <= 1407427199))",\  
-               "steps": [\  
-                 {\  
-                   "transformation": "equality_propagation",\  ---逻辑优化,条件化简,等式处理  
-                   "resulting_condition": "((`t_audit_operate_log`.`Fuser` = ‘XX@XX.com‘) and (`t_audit_operate_log`.`Fcreate_time` >= 1407081600) and (`t_audit_operate_log`.`Fcreate_time` <= 1407427199))"\  
-                 },\  
-                 {\  
-                   "transformation": "constant_propagation",\  ---逻辑优化,条件化简,常量处理  
-                   "resulting_condition": "((`t_audit_operate_log`.`Fuser` = ‘XX@XX.com‘) and (`t_audit_operate_log`.`Fcreate_time` >= 1407081600) and (`t_audit_operate_log`.`Fcreate_time` <= 1407427199))"\  
-                 },\  
-                 {\  
-                   "transformation": "trivial_condition_removal",\  ---逻辑优化,条件化简,条件去除  
-                   "resulting_condition": "((`t_audit_operate_log`.`Fuser` = ‘XX@XX.com‘) and (`t_audit_operate_log`.`Fcreate_time` >= 1407081600) and (`t_audit_operate_log`.`Fcreate_time` <= 1407427199))"\  
-                 }\  
-               ] /* steps */\  
-             } /* condition_processing */\  
-           },\  ---逻辑优化,条件化简,结束  
-           {\  
-             "table_dependencies": [\  ---逻辑优化, 找出表之间的相互依赖关系. 非直接可用的优化方式.   
-               {\  
-                 "table": "`t_audit_operate_log`",\  
-                 "row_may_be_null": false,\  
-                 "map_bit": 0,\  
-                 "depends_on_map_bits": [\  
-                 ] /* depends_on_map_bits */\  
-               }\  
-             ] /* table_dependencies */\  
-           },\  
-           {\  
-             "ref_optimizer_key_uses": [\   ---逻辑优化,  找出备选的索引  
-               {\  
-                 "table": "`t_audit_operate_log`",\  
-                 "field": "Fuser",\  
-                 "equals": "‘XX@XX.com‘",\  
-                 "null_rejecting": false\  
-               }\  
-             ] /* ref_optimizer_key_uses */\  
-           },\  
-           {\  
-             "rows_estimation": [\   ---逻辑优化, 估算每个表的元组个数. 单表上进行全表扫描和索引扫描的代价估算. 每个索引都估算索引扫描代价  
-               {\  
-                 "table": "`t_audit_operate_log`",\  
-                 "range_analysis": {\  
-                   "table_scan": {\---逻辑优化, 估算每个表的元组个数. 单表上进行全表扫描的代价  
-                     "rows": 8150516,\  
-                     "cost": 1.73e6\  
-                   } /* table_scan */,\  
-                   "potential_range_indices": [\ ---逻辑优化, 列出备选的索引. 后续版本字符串变为potential_range_indexes  
-                     {\  
-                       "index": "PRIMARY",\---逻辑优化, 本行表明主键索引不可用  
-                       "usable": false,\  
-                       "cause": "not_applicable"\  
-                     },\  
-                     {\  
-                       "index": "indx_ctime",\---逻辑优化, 索引indx_ctime  
-                       "usable": true,\  
-                       "key_parts": [\  
-                         "Fcreate_time",\  
-                         "Fid"\  
-                       ] /* key_parts */\  
-                     },\  
-                     {\  
-                       "index": "indx_user",\---逻辑优化, 索引indx_user  
-                       "usable": true,\  
-                       "key_parts": [\  
-                         "Fuser",\  
-                         "Fid"\  
-                       ] /* key_parts */\  
-                     },\  
-                     {\  
-                       "index": "indx_objid",\---逻辑优化, 索引  
-                       "usable": false,\  
-                       "cause": "not_applicable"\  
-                     },\  
-                     {\  
-                       "index": "indx_ip",\---逻辑优化, 索引  
-                       "usable": false,\  
-                       "cause": "not_applicable"\  
-                     }\  
-                   ] /* potential_range_indices */,\  
-                   "setup_range_conditions": [\ ---逻辑优化, 如果有可下推的条件,则带条件考虑范围查询  
-                   ] /* setup_range_conditions */,\  
-                   "group_index_range": {\---逻辑优化, 如带有GROUPBY或DISTINCT,则考虑是否有索引可优化这种操作. 并考虑带有MIN/MAX的情况  
-                     "chosen": false,\  
-                     "cause": "not_group_by_or_distinct"\  
-                   } /* group_index_range */,\  
-                   "analyzing_range_alternatives": {\---逻辑优化,开始计算每个索引做范围扫描的花费(等值比较是范围扫描的特例)  
-                     "range_scan_alternatives": [\  
-                       {\  
-                         "index": "indx_ctime",\ ---[A]  
-                         "ranges": [\  
-                           "1407081600 <= Fcreate_time <= 1407427199"\  
-                         ] /* ranges */,\  
-                         "index_dives_for_eq_ranges": true,\  
-                         "rowid_ordered": false,\  
-                         "using_mrr": true,\  
-                         "index_only": false,\  
-                         "rows": 688362,\  
-                         "cost": 564553,\ ---逻辑优化,这个索引的代价最小  
-                         "chosen": true\ ---逻辑优化,这个索引的代价最小,被选中. (比前面的table_scan 和其他索引的代价都小)  
-                       },\  
-                       {\  
-                         "index": "indx_user",\  
-                         "ranges": [\  
-                           "XX@XX.com <= Fuser <= XX@XX.com"\  
-                         ] /* ranges */,\  
-                         "index_dives_for_eq_ranges": true,\  
-                         "rowid_ordered": true,\  
-                         "using_mrr": true,\  
-                         "index_only": false,\  
-                         "rows": 1945894,\  
-                         "cost": 1.18e6,\  
-                         "chosen": false,\  
-                         "cause": "cost"\  
-                       }\  
-                     ] /* range_scan_alternatives */,\  
-                     "analyzing_roworder_intersect": {\  
-                       "usable": false,\  
-                       "cause": "too_few_roworder_scans"\  
-                     } /* analyzing_roworder_intersect */\  
-                   } /* analyzing_range_alternatives */,\---逻辑优化,开始计算每个索引做范围扫描的花费. 这项工作结算  
-                   "chosen_range_access_summary": {\---逻辑优化,开始计算每个索引做范围扫描的花费. 总结本阶段最优的.  
-                     "range_access_plan": {\  
-                       "type": "range_scan",\  
-                       "index": "indx_ctime",\  
-                       "rows": 688362,\  
-                       "ranges": [\  
-                         "1407081600 <= Fcreate_time <= 1407427199"\  
-                       ] /* ranges */\  
-                     } /* range_access_plan */,\  
-                     "rows_for_plan": 688362,\  
-                     "cost_for_plan": 564553,\  
-                     "chosen": true\    -- 这里看到的cost和rows都比 indx_user 要来的小很多---这个和[A]处是一样的,是信息汇总.  
-                   } /* chosen_range_access_summary */\  
-                 } /* range_analysis */\  
-               }\  
-             ] /* rows_estimation */\ ---逻辑优化, 估算每个表的元组个数. 行估算结束  
-           },\  
-           {\  
-             "considered_execution_plans": [\ ---物理优化, 开始多表连接的物理优化计算  
-               {\  
-                 "plan_prefix": [\  
-                 ] /* plan_prefix */,\  
-                 "table": "`t_audit_operate_log`",\  
-                 "best_access_path": {\  
-                   "considered_access_paths": [\  
-                     {\  
-                       "access_type": "ref",\ ---物理优化, 计算indx_user索引上使用ref方查找的花费,  
-                       "index": "indx_user",\  
-                       "rows": 1.95e6,\  
-                       "cost": 683515,\  
-                       "chosen": true\  
-                     },\ ---物理优化, 本应该比较所有的可用索引,即打印出多个格式相同的但索引名不同的内容,这里却没有。推测是bug--没有遍历每一个索引.  
-                     {\  
-                       "access_type": "range",\---物理优化,猜测对应的是indx_time(没有实例可进行调试,对比5.7的跟踪信息猜测而得)  
-                       "rows": 516272,\  
-                       "cost": 702225,\---物理优化,代价大于了ref方式的683515,所以没有被选择  
-                       "chosen": false\   -- cost比上面看到的增加了很多,但rows没什么变化 ---物理优化,此索引没有被选择  
-                     }\  
-                   ] /* considered_access_paths */\  
-                 } /* best_access_path */,\  
-                 "cost_for_plan": 683515,\ ---物理优化,汇总在best_access_path 阶段得到的结果  
-                 "rows_for_plan": 1.95e6,\  
-                 "chosen": true\   -- cost比上面看到的竟然小了很多?虽然rows没啥变化  ---物理优化,汇总在best_access_path 阶段得到的结果  
-               }\  
-             ] /* considered_execution_plans */\  
-           },\  
-           {\  
-             "attaching_conditions_to_tables": {\---逻辑优化,尽量把条件绑定到对应的表上  
-               } /* attaching_conditions_to_tables */\  
-           },\  
-           {\  
-             "refine_plan": [\  
-               {\  
-                 "table": "`t_audit_operate_log`",\---逻辑优化,下推索引条件"pushed_index_condition";其他条件附加到表上做为过滤条件"table_condition_attached"  
-               }\  
-             ] /* refine_plan */\  
-           }\  
-         ] /* steps */\  
-       } /* join_optimization */\ \---逻辑优化和物理优化结束  
-     },\  
-     {\  
-       "join_explain": {} /* join_explain */\  
-     }\  
-   ] /* steps */\  
 
 
 
三 其他一个相似问题
单表扫描,使用ref和range从索引获取数据一例  
http://blog.163.com/li_hx/blog/static/183991413201461853637715/
 
四 问题的解决方式
遇到单表上有多个索引的时候,在MySQL5.6.20版本之前的版本,需要人工强制使用索引,以达到最好的效果.
mysql 执行计划分析三看, explain,profiling,optimizer_trace
标签:inno   const   处事   大量   its   tab   pen   之间   比较   
原文地址:http://www.cnblogs.com/mydriverc/p/7086542.html