问题描述:
MySQL 5.5.15 原sql如下:
select constraint_schema,table_name,constraint_name,constraint_type from information_schema.table_constraints where table_schema not in ('information_schema', 'mysql', 'test',‘performance_schema’);
不只是上面提到的table_constraints,information_schema库下的一下几个表,访问时候都会触发这个“顺手”操作。
information_schema.TABLES
information_schema.STATISTICS
information_schema.PARTITIONS
information_schema.KEY_COLUMN_USAGE
information_schema.TABLE_CONSTRAINTS
information_schema.REFERENTIAL_CONSTRAINTS
show table status . .
show index from ...
当innodb_stats_on_metadata=on 都会触发自动更新统计信息。
问题:
5.6 开始默认innodb_stats_on_metadata=off,why??? 答:为了防止自动更新统计信息在DB高峰时导致BP的swap;查询性能大幅度抖动。
没有定期更新统计信息了么??答:有啊,而且可以是持久化的。
我看到的MySQL 5.5.15 这个版本还是条件是====>
counter > 2000000000 || ((ib_int64_t)counter > 16 + table->stat_n_rows / 16)
下面做了对MySQL 收集统计信息做了扩展:
一.下面针对MySQL community(5.5.15、5.5.36、5.6.16)源代码分析:
1.下面是对MySQL-5.5.15 的源代码的分析:
./storage/innobase/row/row0mysql.c
/*********************************************************************//** Updates the table modification counter and calculates new estimates for table and index statistics if necessary. */ UNIV_INLINE void row_update_statistics_if_needed( /*============================*/ dict_table_t* table) /*!< in: table */ { ulint counter; counter = table->stat_modified_counter; table->stat_modified_counter = counter + 1; /* Calculate new statistics if 1 / 16 of table has been modified since the last time a statistics batch was run, or if stat_modified_counter > 2 000 000 000 (to avoid wrap-around). We calculate statistics at most every 16th round, since we may have a counter table which is very small and updated very often. */ if (counter > 2000000000 || ((ib_int64_t)counter > 16 + table->stat_n_rows / 16)) { dict_update_statistics(table, FALSE /* update even if stats are initialized */); } }
从上面可以看出更新统计信息的条件是:
counter > 2000000000 || ((ib_int64_t)counter > 16 + table->stat_n_rows / 16)
2.下面是对MySQL-5.5.36 的源代码的分析:
---------------------------------------------------------------------------
#通过更新统计信息stat_modified_counter,每个表都有这个表里来维护:
./storage/innobase/row/row0mysql.c
/*********************************************************************//** Updates the table modification counter and calculates new estimates for table and index statistics if necessary. */ UNIV_INLINE void row_update_statistics_if_needed( /*============================*/ dict_table_t* table) /*!< in: table */ { ulint counter; counter = table->stat_modified_counter; table->stat_modified_counter = counter + 1; if (DICT_TABLE_CHANGED_TOO_MUCH(table)) { dict_update_statistics( table, FALSE, /* update even if stats are initialized */ TRUE /* only update if stats changed too much */); } } /*********************************************************************/
规则:每一次DML操作导致1 行更新,stat_modified_counter加1,直到满足更新统计信息的条件,stat_modified_counter的值自动重置为0。
#更新统计信息的条件:(有超过1/16的row被更改过会更新表的条件信息)
./storage/innobase/include/dict0dict.h
/** Calculate new statistics if 1 / 16 of table has been modified since the last time a statistics batch was run. We calculate statistics at most every 16th round, since we may have a counter table which is very small and updated very often. @param t table @return true if the table has changed too much and stats need to be recalculated */ #define DICT_TABLE_CHANGED_TOO_MUCH(t) \ ((ib_int64_t) (t)->stat_modified_counter > 16 + (t)->stat_n_rows / 16) /*********************************************************************/
* 这样有个性能问题,若有多个线程同时检测到阈值,也即是并发调用会多次,,会导致dict_update_statistics函数多次的调用,浪费了系统资源。
解决方法:在dict_update_statistics{}函数对stat_modified_counter加锁,避免并发执行。
#统计新跟更新函数:dict_update_statistics
./storage/innobase/dict/dict0dict.c
/*********************************************************************//** Calculates new estimates for table and index statistics. The statistics are used in query optimization. */ UNIV_INTERN void dict_update_statistics( /*===================*/ dict_table_t* table, /*!< in/out: table */ ibool only_calc_if_missing_stats,/*!< in: only update/recalc the stats if they have not been initialized yet, otherwise do nothing */ ibool only_calc_if_changed_too_much)/*!< in: only update/recalc the stats if the table has been changed too much since the last stats update/recalc */ { dict_index_t* index; ulint sum_of_index_sizes = 0; DBUG_EXECUTE_IF("skip_innodb_statistics", return;); -----------------------------------------------------------------------------
可以优化成:
---------------------------------------------------------------------------
1) 加x锁
2) 索引统计
3) stat_modified_counter 置0
4) 解锁
---------------------------------------------------------------------------
3.下面我们来看下对MySQL 5.6.16 的源代码的分析:
MySQL版本:MySQL 5.6.16-log。
./storage/innobase/row/row0mysql.cc
void row_update_statistics_if_needed( /*============================*/ dict_table_t* table) /*!< in: table */ { ib_uint64_t counter; ib_uint64_t n_rows; if (!table->stat_initialized) { DBUG_EXECUTE_IF( "test_upd_stats_if_needed_not_inited", fprintf(stderr, "test_upd_stats_if_needed_not_inited " "was executed\n"); ); return; } counter = table->stat_modified_counter++; n_rows = dict_table_get_n_rows(table); if (dict_stats_is_persistent_enabled(table)) { if (counter > n_rows / 10 /* 10% */ && dict_stats_auto_recalc_is_enabled(table)) { dict_stats_recalc_pool_add(table); table->stat_modified_counter = 0; } return; } /* Calculate new statistics if 1 / 16 of table has been modified since the last time a statistics batch was run. We calculate statistics at most every 16th round, since we may have a counter table which is very small and updated very often. */ if (counter > 16 + n_rows / 16 /* 6.25% */) { ut_ad(!mutex_own(&dict_sys->mutex)); /* this will reset table->stat_modified_counter to 0 */ dict_stats_update(table, DICT_STATS_RECALC_TRANSIENT); } } /*********************************************************************/
从上面的代码看可以看出:
1.对InnoDB表统计信息持久化时,表的row发生变化大于10%(counter > n_rows / 10 /* 10%)并且innodb_stats_auto_recalc=on,统计信信息会更新(虽然innodb_stats_auto_recalc=on是自动重新计算,但是也是异步的,可能会延时,比如当瞬间的DML批量操作就可能有延时)
2.统计信息非持久化还是和5.5 一致的(表的row发生变化大于1/16时更新统计信息)
3.切记:不能完全依赖于MySQL本身的机制来更新统计信息,线上一些表不及时更新统计信息的我遇到过多次,针对这样的表,我在夜间定期analyze table xxx;
二.MySQL 5.6的改进:
可以配置统计信息的持久化和非持久化(非持久化:5.6之前都是这种)
相关参数:
持久化:
innodb_stats_persistent:on(1)
innodb_stats_persistent_sample_pages:20
非持久化:
innodb_stats_sample_pages:8
相关表:
mysql.innodb_index_stats
mysql.innodb_table_stats
From 5.6.6 开始,统计信息默认是持久化的(即innodb_stats_persistent=on),使用参数innodb_stats_persistent_sample_pages的值,来采样,此时非持久化的参数innodb_stats_sample_pages就无效。
From 5.6.6 开始,使用非持久化的统计信息:
1.set innodb_stats_persistent=0;
2.create|alter table stats_persistent=0;
对单个表开启:
create|alter table...STATS_PERSISTENT [=] {DEFAULT|0|1}
DEFAULT:table的统计信息是否持久化由参数 innodb_stats_persistent 决定。\
总结:From 5.6.6 开始,要么开启统计信息持久化,要么是还用以前的非持久化,二者选一。
参考相关参数:
innodb_stats_method: nulls_equal
, nulls_unequal
, and nulls_ignored
myisam_stats_method:nulls_equal
, nulls_unequal
, and nulls_ignored
--------------------------------------------------------------
基数即value group=N/s (N:表行数 S:average group size)
基数(VG)|值组为不重复的值的个数
nulls_equal:所有的NULL都相等,算作一个值组,这样一旦null值很多的情况下,average group size偏大,导致基数偏小。
nulls_unequal:每一个NULL都相等,算作一个值组,这样一旦null值很多的情况下,如果non-null值组大,而null的值组过多,导致average group size偏小,导致基数偏大,可能导致误走索引
nulls_ignored:所有的null都忽略,不记录索引。
--------------------------------------------------------------
参考:
# http://dev.mysql.com/doc/refman/5.6/en/innodb-parameters.html#sysvar_innodb_stats_method