TiDB 社区智慧合集丨TiDB 相关 SQL 脚本大全

非常感谢各位 TiDBer 在之前 【TiDBer 唠嗑茶话会 48】非正式 TiDB 相关 SQL 脚本征集大赛！( https://asktug.com/t/topic/996635 )里提供的各种常用脚本。

在这篇文章中，我们整理了社区同学提供的一系列 TiDB 相关 SQL 脚本，希望能为大家在 TiDB 的使用过程中提供一些帮助和参考。这些脚本涵盖了常见场景下的 SQL 操作， 欢迎各位 TiDBer 持续补充更新～

未来，我们也将整理更多 TiDB 相关实用指南，帮助大家更好地了解、运用 TiDB，敬请期待！

1 缓存表

贡献者：@ShawnYan

alter table xxx cache|nocache;

2 TSO 时间转换

贡献者：@我是咖啡哥

● 方法一：使用函数 TIDB_PARSE_TSO

SELECT TIDB_PARSE_TSO(437447897305317376);
+------------------------------------+
| TIDB_PARSE_TSO(437447897305317376) |
+------------------------------------+
| 2022-11-18 08:28:17.704000         |
+------------------------------------+
1 row in set (0.25 sec)

● 方法二：使用 pd-ctl

~$ tiup ctl:v6.4.0 pd -i -u http://pdip:2379
Starting component `ctl`: /Users/xxx/.tiup/components/ctl/v6.4.0/ctl pd -i -u http://pdip:2379
>> tso 437447897305317376
system:  2022-11-18 08:28:17.704 +0800 CST
logic:   0

3 读取历史数据

贡献者：@我是咖啡哥

● 使用 AS OF TIMESTAMP 语法读取历史数据，可以通过以下三种方式使用 AS OF TIMESTAMP 语法：

SELECT ... FROM ... AS OF TIMESTAMP

START TRANSACTION READ ONLY AS OF TIMESTAMP

SET TRANSACTION READ ONLY AS OF TIMESTAMP

select * from t as of timestamp '2021-05-26 16:45:26';
start transaction read only as of timestamp '2021-05-26 16:45:26';
set transaction read only as of timestamp '2021-05-26 16:45:26';

● 通过系统变量 tidb_read_staleness 读取历史数据

从 5 秒前至现在的时间范围内选择一个尽可能新的时间戳

set @@tidb_read_staleness="-5";

● 通过系统变量 tidb_snapshot 读取历史数据

设置一个特殊的环境变量，这个是一个 session scope 的变量，其意义为读取这个时间之前的最新的一个版本

set @@tidb_snapshot="2016-10-08 16:45:26";

清空这个变量后，即可读取最新版本数据

set @@tidb_snapshot="";

4 查询 tikv_gc_life_time 和 tikv_gc_safe_point 默认时长

贡献者：@TiDBer_m6V1BalM

select VARIABLE_NAME, VARIABLE_VALUE from mysql.tidb where VARIABLE_NAME like "tikv_gc%";

5 搜索某个用户的 TopN 慢查询

贡献者：@fanruinet

select query_time,query,user
from information_schema.slow_query
where is_internal=false -- 排除 TiDB 内部的慢查询 SQL
and user = "user1" -- 查找的用户名
order by query_time desc
limit 2;

6 统计间隔 5 分钟的数据

贡献者：@forever

SELECT concat(date_format(create_time,'%Y-%m-%d %H:'),floor(date_format(create_time,'%i')/5)),count(*)
FROM jcxx
GROUP BY 1;

7 反解析 digest 成 SQL 文本

贡献者：@hey-hoho

select tidb_decode_sql_digests('["xxxxx"]');

8 不涉及分区表用下面的方式查看表的使用情况

贡献者：@xfworld

select TABLE_SCHEMA,TABLE_NAME,TABLE_ROWS,
(DATA_LENGTH+INDEX_LENGTH)/1024/1024/1024 as table_size from tables order by table_size 
desc limit 20;

9 partition 表提供了分区表和非分区表的资源使用情况

贡献者：@xfworld

select TABLE_SCHEMA,TABLE_NAME,PARTITION_NAME,TABLE_ROWS,
(DATA_LENGTH+INDEX_LENGTH)/1024/1024/1024 as table_size from 
information_schema.PARTITIONS order by table_size desc limit 20;

10 查询分析器中看配置文件参数

贡献者：@Kongdom

show config

SHOW CONFIG 语句用于展示 TiDB 各个组件当前正在应用的配置，请注意，配置与系统变量作用于不同维度，请不要混淆，如果希望获取系统变量信息，请使用 SHOW VARIABLES ( https://docs.pingcap.com/zh/tidb/stable/sql-statement-show-variables ) 语法。

11 查找读流量排名前 10 的热点 region

贡献者：@BraveChen

SELECT DISTINCT region_id 
FROM INFORMATION_SCHEMA.tikv_region_status 
WHERE READ_BYTES > ?
ORDER BY READ_BYTES DESC 
LIMIT 10

12 查看参数和变量的脚本

贡献者：@buddyyuan

#!/bin/bash

case $1 in
-pd)
mysql -uroot -h127.0.0.1 -P4000 -p"" -e "SHOW CONFIG WHERE type ='pd' and name like '%$2%'"
;;
-tidb)
mysql -uroot -h127.0.0.1 -P4000 -p"" -e "SHOW CONFIG WHERE type ='tidb' and name like '%$2%'"
;;
-tikv)
mysql -uroot -h127.0.0.1 -P4000 -p"" -e "SHOW CONFIG WHERE type ='tikv' and name like '%$2%'"
;;
-tiflash)
mysql -uroot -h127.0.0.1 -P4000 -p"" -e "SHOW CONFIG WHERE type ='tiflash' and name like '%$2%'"
;;
-var)
mysql -uroot -h127.0.0.1 -P4000 -p"" -e "show variables like '%$2%';"
;;
-h)
echo "-pd       show pd parameters" 
echo "-tidb     show tidb parameters"
echo "-tikv     show tikv parameters"
echo "-tiflash  show tiflash parameters"
echo "-var      show itidb variables"
;;
esac

还能用 grep 在过滤一次

[root@vm172-16-201-125 ~]# sh showparammeter.sh -tikv memory-pool-quota | grep -i "210:29160"
tikv    192.16.201.210:29160    server.grpc-memory-pool-quota   9223372036854775807B

13 查找重复记录

贡献者：@ealam_ 小羽

select *
from 表
where 重复字段 in
(
select 重复字段
from 表
group by 重复字段
having count(*)>1
)

14 查询耗时最高的慢 sql

贡献者：@caiyfc

select query sql_text,
       sum_query_time,
       mnt as executions,
       avg_query_time,
       avg_proc_time,
       avg_wait_time,
       max_query_time,
       avg_backoff_time,
       Cop_proc_addr,
       digest,
       (case
         when avg_proc_time = 0 then
          'point_get or commit'
         when (avg_proc_time > avg_wait_time and
              avg_proc_time > avg_backoff_time) then
          'coprocessor_process'
         when (avg_backoff_time > avg_wait_time and
              avg_proc_time < avg_backoff_time) then
          'backoff'
         else
          'coprocessor_wait'
       end) as type
  from (select substr(query, 1, 100) query,
               count(*) mnt,
               avg(query_time) avg_query_time,
               avg(process_time) avg_proc_time,
               avg(wait_time) avg_wait_time,
               max(query_time) max_query_time,
               sum(query_time) sum_query_time,
               digest,
               Cop_proc_addr,
               avg(backoff_time) avg_backoff_time
          from information_schema.cluster_slow_query
         where time >= '2022-07-14 17:00:00'
           and time <= '2022-07-15 17:10:00'
           and DB = 'web'
         group by substr(query, 1, 100)) t
 order by max_query_time desc limit 20;

15 日常维护用的最多的 SQL

贡献者：@tracy0984

select * from information_schema.cluster_processlist;

-- kill id;

16 恢复数据（适用于 drop 与 truncate）

贡献者：@凌云 Cloud

FLASHBACK TABLE target_table_name[TO new_table_name]

17 批量修改库名

贡献者：@TiDBer_dog

./bat_rename.sh lihongbao/ dev2_kelun dev2_sinodemo 路径./leo_backup

18 高并发的场景下获取 sql

贡献者：@jiawei

select * from information_schema.processlist where info is not null

19 查看 schema 下的表都有哪些

贡献者：@Ming

show tables in schema;

20 查看表 leader

贡献者：@TiDBer_wTKU9jv6

select count(1),tss.ADDRESS from INFORMATION_SCHEMA.TIKV_REGION_PEERS trp,INFORMATION_SCHEMA.TIKV_REGION_STATUS trs,INFORMATION_SCHEMA.TIKV_STORE_STATUS tss where trp.STORE_ID=tss.STORE_ID and trp.REGION_ID=trs.REGION_ID and trs.DB_NAME='test' and trs.TABLE_NAME='test' and trp.IS_LEADER=1 group by tss.ADDRESS order by tss.ADDRESS;

21 shell 的调皮加速脚本

贡献者：@gcworkerishungry

alias ctidb="mysql -u root -ptidb -Dcktest -h S001 -P4000"

alias dtidb="tiup cluster display tidb-test"

alias etidb="tiup cluster edit-config tidb-test"

alias ptidb="tiup cluster prune tidb-test"

alias rtidb="tiup cluster restart tidb-test"

22 恢复数据到新的数据库

贡献者：@TiDBer_ 徐川

./loader -h 192.168.180.3 -u root -p q1w2 -P 4000 -t 32 -d leo_backup/

23 开启 tiflash

贡献者：@TiDBer_pFFcXLgY

alter table xxx set tiflash replica 1

24 表 region 分布语句

贡献者：@秋枫之舞

select
trs.db_name,
trs.table_name,
trs.index_name,
trp.store_id,
count(*),
sum(approximate_keys)
from
information_schema.tikv_region_status trs,
information_schema.tikv_store_status tss,
information_schema.tikv_region_peers trp
where
trs.db_name = 'prd01'
and trs.table_name = 'tab_name'
and trp.is_leader = 1
and trp.store_id = tss.store_id
and trs.region_id = trp.region_id
group by
trs.db_name,
trs.table_name,
trs.index_name,
trp.store_id
order by
trs.index_name;

25 查看列的元数据

贡献者：@张雨齐0720

show stats_histograms where db_name like 'test' and table_name like 'test1' ;

26 表的存储位置（store、peer 信息）

贡献者：@bert

SELECT distinct a.TIDB_TABLE_ID, b.DB_NAME, b.TABLE_NAME, b.REGION_ID, b.APPROXIMATE_SIZE
, c.PEER_ID, c.STORE_ID, c.IS_LEADER, c.STATUS, d.ADDRESS
, d.STORE_STATE_NAME, d.VERSION, d.CAPACITY, d.AVAILABLE, d.LABEL
FROM INFORMATION_SCHEMA.TABLES a
INNER JOIN INFORMATION_SCHEMA.TIKV_REGION_STATUS b
INNER JOIN INFORMATION_SCHEMA.TIKV_REGION_PEERS c
INNER JOIN INFORMATION_SCHEMA.TIKV_STORE_STATUS d
WHERE a.TIDB_TABLE_ID = b.TABLE_ID
AND b.REGION_ID = c.REGION_ID
AND c.STORE_ID = d.STORE_ID
AND a.TABLE_SCHEMA = 'test'
AND a.TABLE_NAME = 't';

27 将集群升级到指定版本 ( 在线升级 )

贡献者：@TiDBer_ 杨龟干外公

tiup cluster upgrade

例如升级到 v4.0.0 版本：

tiup cluster upgrade tidb-test v4.0.0

28 查询表大小

贡献者：@我是咖啡哥

SELECT
t.TABLE\_NAME,
t.TABLE\_ROWS,
t.TABLE\_TYPE,
round(t.DATA\_LENGTH/1024/1024/1024,2) data\_GB,
round(t.INDEX\_LENGTH/1024/1024/1024,2) index\_GB,
t.CREATE\_OPTIONS,
t.TABLE\_COMMENT
FROM
INFORMATION\_SCHEMA.`TABLES` t
WHERE
table\_schema = 'test'
and t.table\_type='BASE TABLE'
order by t.TABLE\_ROWS desc;

SELECT CONCAT(table\_schema,'.',table\_name) AS 'Table Name', table\_rows AS 'Number of Rows', CONCAT(ROUND(data\_length/(1024*1024*1024),4),'G') AS 'Data Size', CONCAT(ROUND(index\_length/(1024*1024*1024),4),'G') AS 'Index Size', CONCAT(ROUND((data\_length+index\_length)/(1024*1024*1024),4),'G') AS'Total' FROM information\_schema.TABLES WHERE table\_schema LIKE 'test';

29 统计信息

贡献者：@我是咖啡哥

● 查看表的元数据

show stats_meta where db_name like '%sbtest%';

● 查看表的健康状态

show stats\_healthy;

Healthy 字段，一般小于等于 60 的表需要做 analyze

show stats\_healthy where table\_name ='xxx';
show stats\_healthy where db\_name='' and table\_name='orders';
_name like 'sbtest' and table_name like 'sbtest1' ;

● 查看直方图信息

show stats\_buckets where db\_name='' and table\_name='';

● 查看 analyze 状态

show analyze status;

● 分析表、分区

analyze table sbtest1;
ANALYZE TABLE xxx PARTITION P202204;

30 执行计划

贡献者：@我是咖啡哥

绑定执行计划

● 默认是 session 级别

create binding for  select \* from t  using select \* from t use index()

create binding for SELECT  \* FROM t1 INNER JOIN t2 ON t1.id = t2.t1\_id WHERE t1.int\_col = ? using SELECT /\*+ INL\_JOIN(t1, t2) \*/  \* FROM t1 INNER JOIN t2 ON t1.id = t2.t1\_id WHERE t1.int\_col = ?;

explain SELECT  \* FROM t1 INNER JOIN t2 ON t1.id = t2.t1\_id WHERE t1.int\_col = 1;

show bindings for SELECT  \* FROM t1 INNER JOIN t2 ON t1.id = t2.t1\_id WHERE t1.int\_col = 1;

show global bindings;
show session bindings;
SELECT @@SESSION.last\_plan\_from\_binding;

● 使用 explain format = 'verbose' 语句查看 SQL 的执行计划

explain format = 'verbose';

drop binding for sql;

31 查看 regions

贡献者：@我是咖啡哥

SHOW TABLE t\_its\_unload\_priority\_intermediate\_info regions;
SHOW TABLE t\_its\_unload\_priority\_intermediate\_info INDEX IDX\_UPII\_GROUP\_BY\_COMPOSITE regions;

32 统计读写热点表

贡献者：@我是咖啡哥

use INFORMATION\_SCHEMA;

SELECT
db\_name,
table\_name,
index\_name,
type,
sum( flow\_bytes ),
count( 1 ),
group\_concat( h.region\_id ),
count( DISTINCT p.store\_id ),
group\_concat( p.store\_id )
FROM
INFORMATION\_SCHEMA.tidb\_hot\_regions h
JOIN INFORMATION\_SCHEMA.tikv\_region\_peers p ON h.region\_id = p.region\_id
AND p.is\_leader = 1
GROUP BY
db\_name,
table\_name,
index\_name,
type;

SELECT
p.store\_id,
sum(flow\_bytes ),
count(1)
FROM
INFORMATION\_SCHEMA.tidb\_hot\_regions h
JOIN INFORMATION\_SCHEMA.tikv\_region\_peers p ON h.region\_id = p.region\_id
AND p.is\_leader = 1
GROUP BY
p.store\_id
ORDER BY
2 DESC;

select tidb\_decode\_plan();

33 TiFlash

贡献者：@我是咖啡哥

ALTER TABLE t\_test\_time\_type SET TIFLASH REPLICA 1;
SELECT \* FROM information\_schema.tiflash\_replica;

select \* from information\_schema.CLUSTER\_HARDWARE where type='tiflash' and DEVICE\_TYPE='disk' and name='path';

34 admin 命令

贡献者：@我是咖啡哥

admin show ddl jobs;
ADMIN CHECK TABLE t_test;
admin show slow 
ADMIN SHOW TELEMETRY;

35 修改隔离参数

贡献者：@我是咖啡哥

● session 级别修改

Engine 隔离：默认：["tikv", "tidb", "tiflash"] 由于 TiDB Dashboard 等组件需要读取一些存储于 TiDB 内存表区的系统表，因此建议实例级别 engine 配置中始终加入 "tidb" engine。

set session tidb\_isolation\_read\_engines = 'tiflash,tidb';
set @@session.tidb\_isolation\_read\_engines = "tiflash,tidb";

● 手工 Hint

select /\*+ read\_from\_storage(tiflash\[table\_name]) */ ... from table\_name;
select /*+ read\_from\_storage(tiflash\[alias\_a,alias\_b]) \*/ ... from table\_name\_1 as alias\_a, table\_name\_2 as alias\_b where alias\_a.column\_1 = alias\_b.column\_2;

set @@tidb\_allow\_mpp=1;

show  config where name like '%oom%' and type='tidb';

admin show ddl;

36 排错-查看日志

贡献者：@我是咖啡哥

SELECT \* FROM INFORMATION\_SCHEMA.CLUSTER\_LOG t
WHERE time > '2022-08-09 00:00:00' AND time < '2022-08-10 00:00:00'
AND TYPE in ('tikv')
AND `LEVEL` = 'ERROR'
ORDER BY time desc;

37 查询所有节点所在 OS 的 CPU 当前使用率

贡献者：@人如其名

SELECT
  b.time,
  a.hostname,
  a.ip,
  a.types,
  b.cpu_used_percent
FROM
  (
    SELECT
      GROUP_CONCAT(TYPE) AS TYPES,
      SUBSTRING_INDEX(instance, ':', 1) AS ip,
      value AS hostname
    FROM
      information_schema.cluster_systeminfo
    WHERE
      name = 'kernel.hostname'
    GROUP BY
      ip,
      hostname
  ) a,
  (
    SELECT
      time,
      SUBSTRING_INDEX(instance, ':', 1) AS ip,
(100 - value) AS cpu_used_percent
    FROM
      metrics_schema.node_cpu_usage
    WHERE
      MODE = 'idle'
      AND time = NOW()
  ) b
WHERE
  a.ip = b.ip

输出示例：

+----------------------------+-----------------------+----------------+----------------------+--------------------+
| time                       | hostname              | ip             | types                | cpu_used_percent   |
+----------------------------+-----------------------+----------------+----------------------+--------------------+
| 2023-01-10 22:40:15.000000 | localhost.localdomain | 192.168.31.201 | tidb,pd,tikv,tiflash | 11.438079153798114 |
+----------------------------+-----------------------+----------------+----------------------+--------------------+
1 row in set (0.04 sec)

说明：我这里所有类型组件只创建了有一个而且都在一个 os 上，所以只显示了一行。

38 清理 tidb 大量数据的脚本，实现删除百万级别以上的数据，而且不影响 tidb 正常使用

贡献者：@xingzhenxiang

date1=`date --date "7 days ago" +"%Y-%m-%d"`
delete_db_sql="delete from mysql_table where create_date_time<'$date1' limit 10000"
​
i=0
​
while ((++i)); do
a=`/bin/mysql -uroot -p123456 -A mysql_database -h127.0.0.1 --comments -e "${delete_db_sql}" -vvv|grep "Query OK" |awk '{print $3}'`
if(($a<1)); then
break 1
fi
sleep 1
printf "%-4d" $((i))

感谢以上 TiDBer 们贡献的 SQL 脚本～记得点赞收藏，可以随时在你的个人收藏夹里查看到~

未来我们将继续发布更多来自社区的精品内容，希望能为大家提供更多有价值的信息和经验。如果您有任何感兴趣的话题，可以在下方留言，我们会整理相关的资料与大家分享哦！