MYSQL PERFORMANCE_SCHEMA HINTS
ACCOUNTS NOT PROPERLY CLOSING CONNECTIONS [ 1 ]
Works since 5.6
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
;
UNUSED INDEXES [ 2 ]
Works since 5.6
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
;
WHO CREATED TEMPORARY (DISK) TABLES
Works since 5.6
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
;
SELECT schema_name, substr(digest_text, 1, 40) AS statement, 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_digest
WHERE sum_created_tmp_disk_tables > 0
OR sum_created_tmp_tables > 0
;
ACCOUNTS WHICH NEVER CONNECTED SINCE LAST START-UP [ 3 ]
Works since 5.6
SELECT DISTINCT m_u.user, m_u.host
FROM mysql.user m_u
LEFT JOIN performance_schema.accounts ps_a ON m_u.user = ps_a.user AND m_u.host = ps_a.host
WHERE ps_a.user IS NULL
ORDER BY m_u.user
;
USERS WHICH NEVER CONNECTED SINCE LAST START-UP
Works since 5.6
SELECT DISTINCT m_u.user
FROM mysql.user m_u
LEFT JOIN performance_schema.users ps_u ON m_u.user = ps_u.user
WHERE ps_u.user IS NULL
ORDER BY m_u.user
;
TOTALLY UNUSED ACCOUNTS (NEVER CONNECTED SINCE LAST RESTART AND NOT USED TO CHECK STORED PROGRAM OR VIEW PRIVILEGES) SINCE LAST START-UP
Works since 5.6
SELECT DISTINCT m_u.user, m_u.host
FROM mysql.user m_u
LEFT JOIN performance_schema.accounts ps_a ON m_u.user = ps_a.user AND ps_a.host = m_u.host
LEFT JOIN information_schema.views is_v ON is_v.DEFINER = CONCAT(m_u.User, '@', m_u.Host) AND is_v.security_type = 'DEFINER'
LEFT JOIN information_schema.routines is_r ON is_r.DEFINER = CONCAT(m_u.User, '@', m_u.Host) AND is_r.security_type = 'DEFINER'
LEFT JOIN information_schema.events is_e ON is_e.definer = CONCAT(m_u.user, '@', m_u.host)
LEFT JOIN information_schema.triggers is_t ON is_t.definer = CONCAT(m_u.user, '@', m_u.host)
WHERE ps_a.user IS NULL
AND is_v.definer IS NULL
AND is_r.definer IS NULL
AND is_e.definer IS NULL
AND is_t.definer IS NULL
ORDER BY m_u.user, m_u.host
;
SHOW FULL PROCESSLIST
Works since 5.5 (5.1?)
But with filter on Sleep and sorting by time to find the evil query...
SELECT id, user, host, db, command, time, state, LEFT(info, 80) AS info
FROM information_schema.processlist
WHERE command NOT IN ('Sleep', 'Binlog Dump')
ORDER BY time ASC
;
Non blocking version, since 5.6:
SELECT PROCESSLIST_ID AS id, PROCESSLIST_USER AS user, PROCESSLIST_HOST AS host, PROCESSLIST_DB AS db
, PROCESSLIST_COMMAND AS command, PROCESSLIST_TIME AS time, PROCESSLIST_STATE AS state, LEFT(PROCESSLIST_INFO, 80) AS info
FROM performance_schema.threads
WHERE PROCESSLIST_ID IS NOT NULL
AND PROCESSLIST_COMMAND NOT IN ('Sleep', 'Binlog Dump')
ORDER BY PROCESSLIST_TIME ASC
;
STORAGE ENGINES PER SCHEMA
For defining backup strategy, preparing migration to InnoDB or Galera Cluster for MySQL, etc.
Works since 5.5 (5.1?)
SELECT table_schema AS `schema`, engine, COUNT(*) AS `tables`
, ROUND(SUM(data_length)/1024/1024, 0) AS data_mb, ROUND(SUM(index_length)/1024/1024, 0) index_mb
FROM information_schema.tables
WHERE table_schema NOT IN ('mysql', 'information_schema', 'performance_schema')
AND engine IS NOT NULL
GROUP BY table_schema, engine
;
+---------------------+--------+--------+---------+----------+
| schema | engine | tables | data_mb | index_mb |
+---------------------+--------+--------+---------+----------+
| mantis | MyISAM | 31 | 0 | 0 |
| mpm | InnoDB | 3 | 0 | 0 |
| mysql_sequences | InnoDB | 2 | 0 | 0 |
| mysql_sequences | MEMORY | 1 | 0 | 0 |
| otrs | InnoDB | 73 | 13 | 4 |
| quartz | InnoDB | 12 | 0 | 0 |
| tracking | MyISAM | 1 | 0 | 0 |
+---------------------+--------+--------+---------+----------+
TABLES WITHOUT A PRIMARY KEY
Galera Cluster, InnoDB, M/S replication with row based replication does not work well with tables without a Primary Key. To find those the following query helps:
Works since 5.5 (5.1?)
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'
;
+--------------+--------------------+
| table_schema | table_name |
+--------------+--------------------+
| test | t_wo_pk |
| test | t_wo_pk_with_Index |
+--------------+--------------------+
BAD SQL QUERIES OF USERS
Sometimes it could be interesting to find users who do evil SQL Queries which do for examplecreated_tmp_disk_tables
, select_full_join
, select_range_check
or sort_merge_passes
. Those can be found as follows:
SELECT user, host, event_name
, sum_created_tmp_disk_tables AS tmp_disk_tables
, sum_select_full_join AS full_join
, sum_select_range_check AS range_check
, sum_sort_merge_passes AS sort_merge
FROM performance_schema.events_statements_summary_by_account_by_event_name
WHERE sum_created_tmp_disk_tables > 0
OR sum_select_full_join > 0
OR sum_select_range_check > 0
OR sum_sort_merge_passes > 0
ORDER BY sum_sort_merge_passes DESC
LIMIT 10
; +-------+-------------+---------------+-----------------+-----------+-------------+------------+
| user | host | event_name | tmp_disk_tables | full_join | range_check | sort_merge |
+-------+-------------+---------------+-----------------+-----------+-------------+------------+
| user1 | 192.168.0.3 | insert_select | 0 | 7033 | 0 | 10947 |
| user2 | 192.168.0.4 | insert_select | 0 | 6837 | 0 | 10792 |
| user1 | 192.168.0.1 | select | 10742308 | 2095 | 23061 | 16 |
| user2 | 192.168.0.2 | select | 10958067 | 2639 | 23162 | 14 |
+-------+-------------+---------------+-----------------+-----------+-------------+------------+
SHOW PROFILE IN PERFORMANCE_SCHEMA
On systems with heavy traffic the PERFORMANCE_SCHEMA tables might be too small.
mysql> SHOW GLOBAL VARIABLES LIKE 'perf%events%stage%hist%long%';
+----------------------------------------------------+-------+
| Variable_name | Value |
+----------------------------------------------------+-------+
| performance_schema_events_stages_history_long_size | 10000 |
+----------------------------------------------------+-------+ mysql> pager grep history_long
mysql> SHOW ENGINE PERFORMANCE_SCHEMA STATUS;
| performance_schema | events_waits_history_long.row_size | 184 |
| performance_schema | events_waits_history_long.row_count | 1000 |
| performance_schema | events_waits_history_long.memory | 184000 |
| performance_schema | events_stages_history_long.row_size | 88 |
| performance_schema | events_stages_history_long.row_count | 1000 |
| performance_schema | events_stages_history_long.memory | 88000 |
| performance_schema | events_statements_history_long.row_size | 3024 |
| performance_schema | events_statements_history_long.row_count | 1000 |
| performance_schema | events_statements_history_long.memory | 3024000 | mysql> UPDATE performance_schema.setup_instruments
SET ENABLED = 'YES', TIMED = 'YES'
WHERE NAME LIKE '%statement/%';
mysql> UPDATE performance_schema.setup_instruments
SET ENABLED = 'YES', TIMED = 'YES'
WHERE NAME LIKE '%stage/%';
mysql> UPDATE performance_schema.setup_consumers
SET ENABLED = 'YES'
WHERE NAME LIKE '%events_statements_%';
mysql> UPDATE performance_schema.setup_consumers
SET ENABLED = 'YES'
WHERE NAME LIKE '%events_stages_%'; mysql> SELECT ; mysql> SELECT eshl.event_id AS Query_ID, TRUNCATE(eshl.timer_wait/1000000000000, 6) as Duration
, LEFT(eshl.sql_text, 120) AS Query
FROM performance_schema.events_statements_history_long AS eshl
JOIN performance_schema.threads AS t ON t.thread_id = eshl.thread_id
WHERE t.processlist_id = CONNECTION_ID();
+----------+-----------+-------------------------+
| Query_ID | Duration | Query |
+----------+-----------+-------------------------+
| 12 | 13.560737 | select * from test.test |
+----------+-----------+-------------------------+ mysql> SELECT event_name AS Stage, TRUNCATE(timer_wait/1000000000000,6) AS Duration
FROM performance_schema.events_stages_history_long
WHERE nesting_event_id = 12;
+--------------------------------+-----------+
| Stage | Duration |
+--------------------------------+-----------+
| stage/sql/starting | 0.000043 |
| stage/sql/checking permissions | 0.000004 |
| stage/sql/Opening tables | 0.002700 |
| stage/sql/init | 0.000025 |
| stage/sql/System lock | 0.000009 |
| stage/sql/optimizing | 0.000002 |
| stage/sql/statistics | 0.000014 |
| stage/sql/preparing | 0.000013 |
| stage/sql/executing | 0.000000 |
| stage/sql/Sending data | 13.557683 |
| stage/sql/end | 0.000002 |
| stage/sql/query end | 0.000008 |
| stage/sql/closing tables | 0.000006 |
| stage/sql/freeing items | 0.000215 |
| stage/sql/cleaning up | 0.000001 |
+--------------------------------+-----------+
SELECT
, INSERT
, UPDATE
AND DELETE
PER TABLE
Sometimes it is interesting to know how many SELECT
, INSERT
, UPDATE
or DELETE
(DML) statementes have been exectuted against a specifict table (for example for OPTIMZE TABLE
). This can be found as follows:
Works since MySQL 5.6
SELECT object_type, object_schema, object_name
, count_star, count_read, count_write, count_fetch
, count_insert, count_update, count_delete
FROM performance_schema.table_io_waits_summary_by_table
WHERE count_star > 0
; SELECT object_type, object_schema, object_name, index_name
, count_star, count_read, count_write, count_fetch
, count_insert, count_update, count_delete
FROM performance_schema.table_io_waits_summary_by_index_usage
WHERE count_star > 0
;
TOP LONG RUNNING QUERIES
Works since MySQL 5.6
UPDATE setup_consumers SET enabled = 1 WHERE name = 'events_statements_history_long'; SELECT left(digest_text, 64)
, ROUND(SUM(timer_end-timer_start)/1000000000, 1) AS tot_exec_ms
, ROUND(SUM(timer_end-timer_start)/1000000000/COUNT(*), 1) AS avg_exec_ms
, ROUND(MIN(timer_end-timer_start)/1000000000, 1) AS min_exec_ms
, ROUND(MAX(timer_end-timer_start)/1000000000, 1) AS max_exec_ms
, ROUND(SUM(timer_wait)/1000000000, 1) AS tot_wait_ms
, ROUND(SUM(timer_wait)/1000000000/COUNT(*), 1) AS avg_wait_ms
, ROUND(MIN(timer_wait)/1000000000, 1) AS min_wait_ms
, ROUND(MAX(timer_wait)/1000000000, 1) AS max_wait_ms
, ROUND(SUM(lock_time)/1000000000, 1) AS tot_lock_ms
, ROUND(SUM(lock_time)/1000000000/COUNT(*), 1) AS avglock_ms
, ROUND(MIN(lock_time)/1000000000, 1) AS min_lock_ms
, ROUND(MAX(lock_time)/1000000000, 1) AS max_lock_ms
, MIN(LEFT(DATE_SUB(NOW(), INTERVAL (isgs.VARIABLE_VALUE - TIMER_START*10e-13) second), 19)) AS first_seen
, MAX(LEFT(DATE_SUB(NOW(), INTERVAL (isgs.VARIABLE_VALUE - TIMER_START*10e-13) second), 19)) AS last_seen
, COUNT(*) as cnt
FROM events_statements_history_long
JOIN information_schema.global_status AS isgs
WHERE isgs.variable_name = 'UPTIME'
GROUP BY LEFT(digest_text,64)
ORDER BY tot_exec_ms DESC
; +------------------------------------------------------------------+-------------+-------------+-------------+---------------------+---------------------+-----+
| left(digest_text, 64) | tot_exec_ms | tot_wait_ms | tot_lock_ms | first_seen | last_seen | cnt |
+------------------------------------------------------------------+-------------+-------------+-------------+---------------------+---------------------+-----+
| INSERT INTO `test` SELECT ? , DATA , ? FROM `test` | 50493.5 | 50493.5 | 26.3 | 2015-11-12 16:41:35 | 2015-11-12 16:42:04 | 20 |
| SELECT LEFT ( `digest_text` , ? ) , `ROUND` ( SUM ( `timer_end` | 14434.6 | 14434.6 | 25.8 | 2015-11-12 16:48:44 | 2015-11-12 17:07:15 | 6 |
| SELECT * FROM `test` | 7483.0 | 7483.0 | 0.2 | 2015-11-12 16:41:16 | 2015-11-12 16:42:34 | 2 |
| SHOW ENGINE INNODB STATUS | 1912.4 | 1912.4 | 0.0 | 2015-11-12 16:37:19 | 2015-11-12 17:07:36 | 687 |
| SHOW GLOBAL VARIABLES | 1091.1 | 1091.1 | 68.8 | 2015-11-12 16:37:19 | 2015-11-12 17:07:36 | 687 |
| SHOW GLOBAL STATUS | 638.7 | 638.7 | 40.8 | 2015-11-12 16:37:19 | 2015-11-12 17:07:36 | 687 |
| SELECT LEFT ( `digest_text` , ? ) , SUM ( `timer_end` - `timer_s | 356.2 | 356.2 | 42.4 | 2015-11-12 16:42:38 | 2015-11-12 16:45:00 | 6 |
| SELECT `digest_text` , SUM ( `timer_end` - `timer_start` ) / ? A | 325.3 | 325.3 | 0.4 | 2015-11-12 16:40:44 | 2015-11-12 16:42:18 | 3 |
| SELECT `DIGEST_TEXT` , ( `TIMER_END` - `TIMER_START` ) / ? AS `e | 163.2 | 163.2 | 1.0 | 2015-11-12 16:37:44 | 2015-11-12 16:39:22 | 9 |
| SELECT LOWER ( REPLACE ( trx_state , ?, ... ) ) AS state , COUNT | 133.9 | 133.9 | 80.2 | 2015-11-12 16:37:19 | 2015-11-12 17:07:36 | 687 |
+------------------------------------------------------------------+-------------+-------------+-------------+---------------------+---------------------+-----+
TABLES NEVER WRITTEN TO
If you want to find tables which it was never written to (or read from) since last instance restart you can use the following query. Works since MySQL 5.6
Caution:
count_read
can only be taken as argument if a physical backup method (NOTmysqldump
) is used.- If
count_write
is 0 it does not necessarily mean that there was no write statement (no matching write)! - If tables are empty
SELECT
statements are not counted (count_read
= 0).
SELECT t.table_schema, t.table_name, t.table_rows, tio.count_read, tio.count_write
FROM information_schema.tables AS t
JOIN performance_schema.table_io_waits_summary_by_table AS tio
ON tio.object_schema = t.table_schema AND tio.object_name = t.table_name
WHERE t.table_schema NOT IN ('mysql', 'information_schema', 'performance_schema', 'sys')
AND tio.count_write = 0
ORDER BY t.table_schema, t.table_name
;
http://www.fromdual.com/mysql-performance-schema-hints
MYSQL PERFORMANCE_SCHEMA HINTS的更多相关文章
- mysql performance_schema 初探
mysql performance_schema 初探: mysql 5.5 版本 新增了一个性能优化的引擎: PERFORMANCE_SCHEMA 这个功能默认是关闭的: 需要设置参数: perf ...
- mysql performance_schema 和information_schema.tables了解
这个是关于mysql的系统表,性能表,核心表操作的一些介绍,深入算不上 我们一般很少去动 mysql information_schema 信息相关 performance_schema 性能相关 ...
- mysql performance_schema/information_schema授权问题
mysql> grant all on performance_schema.* to 'testuser'@'%';ERROR 1044 (42000): Access denied for ...
- docker配置mysql主从与django实现读写分离
一.搭建主从mysql环境 1 下载mysql镜像 docker pull mysql:5.7 2 运行刚下载的mysql镜像文件 # 运行该命令之前可以使用`docker images`是否下载成功 ...
- 基于MySQL MEB的备份恢复
MEB(MySQL Enterprise Backup)是MySQL商业版中提供的备份工具,属于物理备份. 同XtraBackup一样,mysqlbackup的使用过程同样包含如下三个步骤: 备份(- ...
- 如何在删除ibdata1和ib_logfile的情况下恢复MySQL数据库
昨天,有个朋友对公司内部使用的一个MySQL实例开启binlog,但是在启动的过程中失败了(他也没提,为何会失败),在启动失败后,他删除了ibdata1和ib_logfile,后来,能正常启动了,但所 ...
- 读书笔记--SQL必知必会--常用MySQL(MariaDB)命令
DBMS信息 显示DBMS的版本 select version(); 显示DBMS状态 status; 显示DBMS资源状态 show status; 显示DBMS支持的权限 show privile ...
- 4.MySQL 主主(m-m) 同步生产库标准同步操作实施流程
通过MySQL参数配置使用主主前提: 1.表的主键自增. ################################################################# #m1-m ...
- 二进制包安装MySQL数据库
1.1二进制包安装MySQL数据库 1.1.1 安装前准备(规范) [root@Mysql_server ~]# mkdir -p /home/zhurui/tools ##创建指定工具包存放路径 [ ...
随机推荐
- iOS RSA加密解密及签名验证
1.首先要下载openssl,这个不用说,直接官网下载或者用brew install openssl下载 2.终端生成私钥密钥 2.1生成私钥 openssl genrsa - 2.2生成密钥 ope ...
- 用.NET从外部dwg文件导入块
翻译并引自Kean's blog的两篇文章: http://through-the-interface.typepad.com/through_the_interface/2006/08/import ...
- 记录一些PHP7RCC1编译问题
1,php7rc1源码编译undefined symboles的问题 自己计划将php7环境部署到cubieboard上,懒得去找别人预编译的版本,所以动手从源码编译,中间遇到了一个小问题,此处记录一 ...
- Python通用序列操作
1.序列概览 1.数据结构 序列.容器 Python中最基本的数据结构是序列,其有索引(从左到右第一个索引为0,从右到左第一个索引为-1). Python包含6中内建的序列: 列表 元组 字符串 Un ...
- Python数学函数
1.Python数学函数 1.abs(x):取绝对值,内建函数 2.math.ceil(x):向上取整,在math模块中 3.cmp(x,y):如果 x < y ,返回-1:如果 x == y ...
- Linux中Main函数的执行过程
1. 问题:Linux如何执行main函数. 本文使用一个简单的C程序(simple.c)作为例子讲解.代码如下, int main() { return(0); } 2. 编译 -#gcc -o ...
- 配置不当导致无法加载odoo-10.0模块
启动odoo-bin时出错 2017-01-05 06:38:51,046 5480 INFO ? odoo: Odoo version 10.02017-01-05 06:38:51,046 548 ...
- js数组的队栈操作
<script> //一个数组 var testArray = [1,2,3,4,5]; //shift()方法,返回并移除数组中的第一个元素的值 testArray.shift();// ...
- HTTP状态代码含义
1.200:正常. 2.201:服务器已经创建了文档,Location头给出了它的URL. 3.202:已经接受请求,但处理尚未完成. 4.203:文档已经正常返回,但一些应答头可能不正确,因为使用的 ...
- compile vim with lua & python support
vim在macosx 10.9默认没有带lua和python支持,因为装的有些插件是lua写的,有些是python写的,运行不起来,于是决定自己编译一个,下载vim源码,执行以下命令就可以编译vim: ...