mysql sys table

本文详细地介绍了MySQL 5.7新引入的sys schema。首先，本文概要地介绍了sys schema的作用和定位；其次，分别介绍了sys schema中的视图、函数和存储过程；接下来，通过两个例子来演示sys schema的用法，便于大家理解sys schema带来的实实在在的好处；最后讨论了sys schema还可以增加的内容。
1. sys schema的介绍

sys schema是MySQL 5.7.7中引入的一个系统库，包含了一系列视图、函数和存储过程， 该项目专注于MySQL的易用性，例如，我们可以通过sys schema快速的知道，哪些语句使用了临时表，哪个用户请求了最多的io，哪个线程占用了最多的内存，哪些索引是无用索引等。

引入sys schema以后，MySQL的易用性将会得到极大地提升，MySQL的用户分析问题和定位问题，将更多的依赖sys schema，减少外部工具的使用。

前面说过，sys schema中包含了大量的视图（只有sys_config是innodb表），那么，这些视图的信息来自哪里呢？视图中的信息均来自performance schema和information schema中的统计信息。MySQL Server blog中有一个很好的比喻：

For Linux users I like to compare performance_schema to /proc, and SYS to vmstat.

也就是说，performance schema和information schema中提供了信息源，但是，没有很好的将这些信息组织成有用的信息，从而没有很好的发挥它们的作用。而sys schema使用performance schema和information schema中的信息，通过视图的方式给出解决实际问题的答案。这就是sys schema的作用和目的，也是为什么sys schema值得我们花点时间学习的原因。
2. sys schema中的视图、函数和存储过程

可以通过以下语句快速查看sys schema包含的视图、函数和存储过程

show full tables from sys
show function status where db = ‘sys‘;
show procedure status where db = ‘sys‘

接下来将依次给出所有的视图、函数和存储过程，并进行简单的分析，希望能够达到抛砖引玉的效果。
2.1 视图

sys schema中的视图（和一张表）如下，通过名称就很容易猜到具体是做什么用的。

mysql> select table_name , table_type, engine from information_schema.tables where table_schema = ‘sys‘ order by table_name;
+-----------------------------------------------+------------+--------+
| table_name | table_type | engine |
+-----------------------------------------------+------------+--------+
| host_summary | VIEW | NULL |
| host_summary_by_file_io | VIEW | NULL |
| host_summary_by_file_io_type | VIEW | NULL |
| host_summary_by_stages | VIEW | NULL |
| host_summary_by_statement_latency | VIEW | NULL |
| host_summary_by_statement_type | VIEW | NULL |
| innodb_buffer_stats_by_schema | VIEW | NULL |
| innodb_buffer_stats_by_table | VIEW | NULL |
| innodb_lock_waits | VIEW | NULL |
| io_by_thread_by_latency | VIEW | NULL |
| io_global_by_file_by_bytes | VIEW | NULL |
| io_global_by_file_by_latency | VIEW | NULL |
| io_global_by_wait_by_bytes | VIEW | NULL |
| io_global_by_wait_by_latency | VIEW | NULL |
| latest_file_io | VIEW | NULL |
| memory_by_host_by_current_bytes | VIEW | NULL |
| memory_by_thread_by_current_bytes | VIEW | NULL |
| memory_by_user_by_current_bytes | VIEW | NULL |
| memory_global_by_current_bytes | VIEW | NULL |
| memory_global_total | VIEW | NULL |
| metrics | VIEW | NULL |
| processlist | VIEW | NULL |
| ps_check_lost_instrumentation | VIEW | NULL |
| schema_auto_increment_columns | VIEW | NULL |
| schema_index_statistics | VIEW | NULL |
| schema_object_overview | VIEW | NULL |
| schema_redundant_indexes | VIEW | NULL |
| schema_tables_with_full_table_scans | VIEW | NULL |
| schema_table_lock_waits | VIEW | NULL |
| schema_table_statistics | VIEW | NULL |
| schema_table_statistics_with_buffer | VIEW | NULL |
| schema_unused_indexes | VIEW | NULL |
| session | VIEW | NULL |
| session_ssl_status | VIEW | NULL |
| statements_with_errors_or_warnings | VIEW | NULL |
| statements_with_full_table_scans | VIEW | NULL |
| statements_with_runtimes_in_95th_percentile | VIEW | NULL |
| statements_with_sorting | VIEW | NULL |
| statements_with_temp_tables | VIEW | NULL |
| statement_analysis | VIEW | NULL |
| sys_config | BASE TABLE | InnoDB |
| user_summary | VIEW | NULL |
| user_summary_by_file_io | VIEW | NULL |
| user_summary_by_file_io_type | VIEW | NULL |
| user_summary_by_stages | VIEW | NULL |
| user_summary_by_statement_latency | VIEW | NULL |
| user_summary_by_statement_type | VIEW | NULL |
| version | VIEW | NULL |
| waits_by_host_by_latency | VIEW | NULL |
| waits_by_user_by_latency | VIEW | NULL |
| waits_global_by_latency | VIEW | NULL |
| wait_classes_global_by_avg_latency | VIEW | NULL |
| wait_classes_global_by_latency | VIEW | NULL |

这里还没有列出全部的视图，对于每一个视图，都有两种形式：

mysql> select table_name from information_schema.tables where table_schema = ‘sys‘ and table_name like ‘%memory%‘;
+-------------------------------------+
| table_name |
+-------------------------------------+
| memory_by_host_by_current_bytes |
| memory_by_thread_by_current_bytes |
| memory_by_user_by_current_bytes |
| memory_global_by_current_bytes |
| memory_global_total |
| x$memory_by_host_by_current_bytes |
| x$memory_by_thread_by_current_bytes |
| x$memory_by_user_by_current_bytes |
| x$memory_global_by_current_bytes |
| x$memory_global_total |
+-------------------------------------+
10 rows in set (0.00 sec)

前一种是便于人类阅读的形式，格式化了时间单位和字节单位，另外一种形式是指以”x$”开头的视图名称，是为了便于工具处理。如下所示：

mysql> select * from host_summary_by_file_io;
+-------------+------+------------+
| host | ios | io_latency |
+-------------+------+------------+
| background | 2143 | 266.48 ms |
| 172.17.42.1 | 1748 | 116.52 ms |
+-------------+------+------------+
2 rows in set (0.00 sec)

mysql> select * from x$host_summary_by_file_io;
+-------------+------+--------------+
| host | ios | io_latency |
+-------------+------+--------------+
| background | 2148 | 266558291670 |
| 172.17.42.1 | 1748 | 116518395300 |
+-------------+------+--------------+
2 rows in set (0.00 sec)

我们再看一下host_summary_by_file_io的定义，如下所示：

mysql> show create table host_summary_by_file_io\G
*************************** 1. row ***************************
View: host_summary_by_file_io
Create View: CREATE ALGORITHM=TEMPTABLE DEFINER=`mysql.sys`@`localhost` SQL SECURITY INVOKER VIEW `host_summary_by_file_io` AS select if(isnull(`performance_schema`.`events_waits_summary_by_host_by_event_name`.`HOST`),‘background‘,`performance_schema`.`events_waits_summary_by_host_by_event_name`.`HOST`) AS `host`,sum(`performance_schema`.`events_waits_summary_by_host_by_event_name`.`COUNT_STAR`) AS `ios`,`sys`.`format_time`(sum(`performance_schema`.`events_waits_summary_by_host_by_event_name`.`SUM_TIMER_WAIT`)) AS `io_latency` from `performance_schema`.`events_waits_summary_by_host_by_event_name` where (`performance_schema`.`events_waits_summary_by_host_by_event_name`.`EVENT_NAME` like ‘wait/io/file/%‘) group by if(isnull(`performance_schema`.`events_waits_summary_by_host_by_event_name`.`HOST`),‘background‘,`performance_schema`.`events_waits_summary_by_host_by_event_name`.`HOST`) order by sum(`performance_schema`.`events_waits_summary_by_host_by_event_name`.`SUM_TIMER_WAIT`) desc
character_set_client: utf8
collation_connection: utf8_general_ci
1 row in set (0.00 sec)

可以看到，定义非常复杂，如果不是sys schema提供了这么好用的视图，我们自己几乎无法从这么多信息中获取到有用的信息。
2.2 函数

如果你仔细看了上面的定义视图的SQL语句，你可能注意它使用了一个叫做sys.format_time的函数，这个函数也是随着sys schema一起发布，用以格式化时间单位。sys schema还包含了其他一些有用的函数，如下所示：

mysql> select ROUTINE_SCHEMA, ROUTINE_NAME, ROUTINE_TYPE from ROUTINES where ROUTINE_TYPE=‘FUNCTION‘;
+----------------+----------------------------------+--------------+
| ROUTINE_SCHEMA | ROUTINE_NAME | ROUTINE_TYPE |
+----------------+----------------------------------+--------------+
| sys | extract_schema_from_file_name | FUNCTION |
| sys | extract_table_from_file_name | FUNCTION |
| sys | format_bytes | FUNCTION |
| sys | format_path | FUNCTION |
| sys | format_statement | FUNCTION |
| sys | format_time | FUNCTION |
| sys | list_add | FUNCTION |
| sys | list_drop | FUNCTION |
| sys | ps_is_account_enabled | FUNCTION |
| sys | ps_is_consumer_enabled | FUNCTION |
| sys | ps_is_instrument_default_enabled | FUNCTION |
| sys | ps_is_instrument_default_timed | FUNCTION |
| sys | ps_is_thread_instrumented | FUNCTION |
| sys | ps_thread_account | FUNCTION |
| sys | ps_thread_id | FUNCTION |
| sys | ps_thread_stack | FUNCTION |
| sys | ps_thread_trx_info | FUNCTION |
| sys | sys_get_config | FUNCTION |
| sys | version_major | FUNCTION |
| sys | version_minor | FUNCTION |
| sys | version_patch | FUNCTION |
+----------------+----------------------------------+--------------+
21 rows in set (0.00 sec)

2.3 存储过程

下面的是sys schema包含的存储过程。可以看到，大部分存储过程都是PS开头的，PS是performance schema的简写，这些存储过程都是用来控制performance schema统计的行为。例如，ps_truncate_all_tables 的作用就是truncate所有performance schema中summary相关的表。

如下所示：

mysql> select ROUTINE_SCHEMA, ROUTINE_NAME, ROUTINE_TYPE from ROUTINES where ROUTINE_TYPE=‘PROCEDURE‘;
+----------------+-------------------------------------+--------------+
| ROUTINE_SCHEMA | ROUTINE_NAME | ROUTINE_TYPE |
+----------------+-------------------------------------+--------------+
| sys | create_synonym_db | PROCEDURE |
| sys | diagnostics | PROCEDURE |
| sys | execute_prepared_stmt | PROCEDURE |
| sys | ps_setup_disable_background_threads | PROCEDURE |
| sys | ps_setup_disable_consumer | PROCEDURE |
| sys | ps_setup_disable_instrument | PROCEDURE |
| sys | ps_setup_disable_thread | PROCEDURE |
| sys | ps_setup_enable_background_threads | PROCEDURE |
| sys | ps_setup_enable_consumer | PROCEDURE |
| sys | ps_setup_enable_instrument | PROCEDURE |
| sys | ps_setup_enable_thread | PROCEDURE |
| sys | ps_setup_reload_saved | PROCEDURE |
| sys | ps_setup_reset_to_default | PROCEDURE |
| sys | ps_setup_save | PROCEDURE |
| sys | ps_setup_show_disabled | PROCEDURE |
| sys | ps_setup_show_disabled_consumers | PROCEDURE |
| sys | ps_setup_show_disabled_instruments | PROCEDURE |
| sys | ps_setup_show_enabled | PROCEDURE |
| sys | ps_setup_show_enabled_consumers | PROCEDURE |
| sys | ps_setup_show_enabled_instruments | PROCEDURE |
| sys | ps_statement_avg_latency_histogram | PROCEDURE |
| sys | ps_trace_statement_digest | PROCEDURE |
| sys | ps_trace_thread | PROCEDURE |
| sys | ps_truncate_all_tables | PROCEDURE |
| sys | statement_performance_analyzer | PROCEDURE |
| sys | table_exists | PROCEDURE |
+----------------+-------------------------------------+--------------+
26 rows in set (0.00 sec)

3. sys schema示例

这一节简单的演示sys schema的用法，便于大家理解sys schema带来的实实在在的好处，然后讨论了sys schema还可以增加的内容。
3.1 sys schema的用法

我们来看几个与索引相关的例子：

如何查看数据库中的冗余索引，在MySQL 5.7之前，需要使用percona的 pt-duplicate-key-checker，现在直接查询sys.schema_redundant_indexes就可以了:

select * from sys.schema_redundant_indexes;

如何获取未使用的索引

在MySQL 5.7之前，我们也可以通过information schema查看未使用的索引：

SELECT object_schema, object_name, index_name
FROM performance_schema.table_io_waits_summary_by_index_usage
WHERE index_name IS NOT NULL
AND count_star = 0
ORDER BY object_schema, object_name ;

有了sys schema以后，直接查询schema_unused_indexes即可:

select * from schema_unused_indexes;

如何查看使用全表扫描的SQL语句（MySQL 5.7之前我不知道怎么看额(⊙o⊙)…） 有了sys schema以后，直接查询statements_with_full_table_scans即可：

select * from statements_with_full_table_scans

3.2 sys schema可以引入的视图

我发现，还有一些非常有用的信息，可以通过information schema 和performance schema获取到，但是，在sys schema中却没有。有鉴于此，个人感觉未来sys schema中的视图会继续增加。例如：

查看没有主键的表

SELECT DISTINCT t.table_schema, t.table_name
FROM information_schema.tables AS t
LEFT JOIN information_schema.columns AS c ON t.table_schema = c.table_schema AND t.table_name = c.table_name AND c.column_key = "PRI"
WHERE t.table_schema NOT IN (‘information_schema‘, ‘mysql‘, ‘performance_schema‘)
AND c.table_name IS NULL AND t.table_type != ‘VIEW‘;

查看是谁创建的临时表

SELECT user, host, event_name, count_star AS cnt, sum_created_tmp_disk_tables AS tmp_disk_tables, sum_created_tmp_tables AS tmp_tables
FROM performance_schema.events_statements_summary_by_account_by_event_name
WHERE sum_created_tmp_disk_tables > 0
OR sum_created_tmp_tables > 0 ;

没有正确关闭数据库连接的用户

SELECT ess.user, ess.host
, (a.total_connections - a.current_connections) - ess.count_star as not_closed
, ((a.total_connections - a.current_connections) - ess.count_star) * 100 /
(a.total_connections - a.current_connections) as pct_not_closed
FROM performance_schema.events_statements_summary_by_account_by_event_name ess
JOIN performance_schema.accounts a on (ess.user = a.user and ess.host = a.host)
WHERE ess.event_name = ‘statement/com/quit‘
AND (a.total_connections - a.current_connections) > ess.count_star ;

这些都是比较有用的信息，适合加入到sys schema中。
4. 总结

本文详细地介绍了sys schema的作用和组成，并通过几个例子演示了sys schema的易用性和不足，相信通过这篇文章，大家对sys schema能够有一个较完整的了解，也愿意花时间去深入了解sys schema中的各个视图。

此外，sys schema可以应用在MySQL 5.6上，具体信息见这里。