postgresql 定时收集表和索引统计信息

本文涉及的产品
云原生数据库 PolarDB MySQL 版,通用型 2核4GB 50GB
云原生数据库 PolarDB PostgreSQL 版,标准版 2核4GB 50GB
简介:

--由于pg中表和索引的信息收集都是基于时间点的,对于以往的信息无法与现在的信息进行对比,故写下此工具进行统计信息收集

--创建数据信息的schema
create schema db_stat;
--创建收集信息的基础表
create table db_stat.snapshot_pg_stat_all_indexes
(relid int,indexrelid int,schemaname varchar(200),relname varchar(550),indexrelname varchar(550),idx_scan bigint,idx_tup_read bigint,idx_tup_fetch bigint,
snapid int,snap_create_time timestamp,host_ip cidr,host_port int,host_type varchar(20),comment varchar(500));
create index idx_stat_indexe_relname_indexrelname_createtime on db_stat.snapshot_pg_stat_all_indexes(relname varchar_pattern_ops,indexrelname varchar_pattern_ops,snap_create_time);
create index idx_stat_indexe_snapid on db_stat.snapshot_pg_stat_all_indexes(snapid);

create table db_stat.snapshot_pg_stat_all_tables
(relid int,schemaname varchar(200),relname varchar(550),seq_scan bigint,seq_tup_read bigint,idx_scan bigint,idx_tup_fetch bigint,n_tup_ins bigint,n_tup_upd bigint,
n_tup_del bigint,n_tup_hot_upd bigint,n_live_tup bigint,n_dead_tup bigint,last_vacuum timestamp,last_autovacuum timestamp,last_analyze timestamp,last_autoanalyze timestamp,vacuum_count bigint,autovacuum_count bigint,analyze_count bigint,autoanalyze_count bigint,
snapid int,snap_create_time timestamp,host_ip cidr,host_port int,host_type varchar(20),comment varchar(500));
create index idx_stat_table_relname_createtime on db_stat.snapshot_pg_stat_all_tables(relname varchar_pattern_ops,snap_create_time);
create index idx_stat_table_snapid on db_stat.snapshot_pg_stat_all_tables(snapid);

create table db_stat.snapshot_pg_statio_all_indexes
(relid int,indexrelid int,schemaname varchar(200),relname varchar(550),indexrelname varchar(550),idx_blks_read bigint,idx_blks_hit bigint,
snapid int,snap_create_time timestamp,host_ip cidr,host_port int,host_type varchar(20),comment varchar(500));
create index idx_statio_indexe_relname_indexrelname_createtime on db_stat.snapshot_pg_statio_all_indexes(relname varchar_pattern_ops,indexrelname varchar_pattern_ops,snap_create_time);
create index idx_statio_indexe_snapid on db_stat.snapshot_pg_statio_all_indexes(snapid);
create table db_stat.snapshot_pg_statio_all_tables
(relid int,schemaname varchar(200),relname varchar(550),heap_blks_read bigint,heap_blks_hit bigint,idx_blks_read bigint,idx_blks_hit bigint,toast_blks_read bigint,toast_blks_hit bigint,
tidx_blks_read bigint,tidx_blks_hit bigint,
snapid int,snap_create_time timestamp,host_ip cidr,host_port int,host_type varchar(20),comment varchar(500));
create index idx_statio_table_relname_createtime on db_stat.snapshot_pg_statio_all_tables(relname varchar_pattern_ops,snap_create_time);
create index idx_statio_table_snapid on db_stat.snapshot_pg_statio_all_tables(snapid);


--创建快照的序列
create sequence db_stat.seq_snapshot minvalue 1 maxvalue 99999999999999;

--每收集完信息之后,对时间,主机列等进行填充
create or replace function db_stat.process_snapshot_table(in i_host_ip cidr,in i_host_port int,in i_host_type varchar,in i_comment varchar default '') returns int as $$
declare
v_snapid int;
_detail text;
_hint text;
_message text;
begin
select nextval('db_stat.seq_snapshot') into v_snapid;
update db_stat.snapshot_pg_stat_all_indexes set snapid=v_snapid,snap_create_time=now(),host_ip=i_host_ip,host_port=i_host_port,host_type=i_host_type,comment=i_comment where snapid is null;
update db_stat.snapshot_pg_stat_all_tables set snapid=v_snapid,snap_create_time=now(),host_ip=i_host_ip,host_port=i_host_port,host_type=i_host_type,comment=i_comment where snapid is null;
update db_stat.snapshot_pg_statio_all_indexes set snapid=v_snapid,snap_create_time=now(),host_ip=i_host_ip,host_port=i_host_port,host_type=i_host_type,comment=i_comment where snapid is null;
update db_stat.snapshot_pg_statio_all_tables set snapid=v_snapid,snap_create_time=now(),host_ip=i_host_ip,host_port=i_host_port,host_type=i_host_type,comment=i_comment where snapid is null;
-- 返回值 1 代表成功,0 代表失败
return 1;
EXCEPTION WHEN others then
GET STACKED DIAGNOSTICS
_message = message_text,
_detail = pg_exception_detail,
_hint = pg_exception_hint;
raise notice 'message: %, detail: %, hint: %', _message, _detail, _hint;
return 0;
end;  
$$ language plpgsql;


--收动进行信息采集,测试用
INSERT INTO db_stat.snapshot_pg_stat_all_indexes(relid ,indexrelid ,schemaname ,relname ,indexrelname,idx_scan ,idx_tup_read,idx_tup_fetch)
SELECT relid ,
       indexrelid ,
       schemaname ,
       relname ,
       indexrelname,
       idx_scan ,
       idx_tup_read,
       idx_tup_fetch
FROM pg_stat_all_indexes;

INSERT INTO db_stat.snapshot_pg_stat_all_tables(relid ,schemaname ,relname ,seq_scan ,seq_tup_read ,idx_scan ,idx_tup_fetch ,n_tup_ins ,n_tup_upd ,n_tup_del ,n_tup_hot_upd ,n_live_tup ,n_dead_tup ,last_vacuum ,last_autovacuum ,last_analyze ,last_autoanalyze ,vacuum_count ,autovacuum_count ,analyze_count ,autoanalyze_count)
SELECT relid ,
       schemaname ,
       relname ,
       seq_scan ,
       seq_tup_read ,
       idx_scan ,
       idx_tup_fetch ,
       n_tup_ins ,
       n_tup_upd ,
       n_tup_del ,
       n_tup_hot_upd ,
       n_live_tup ,
       n_dead_tup ,
       last_vacuum ,
       last_autovacuum ,
       last_analyze ,
       last_autoanalyze ,
       vacuum_count ,
       autovacuum_count ,
       analyze_count ,
       autoanalyze_count
FROM pg_stat_all_tables;


INSERT INTO db_stat.snapshot_pg_statio_all_indexes(relid ,indexrelid ,schemaname ,relname ,indexrelname ,idx_blks_read ,idx_blks_hit)
SELECT relid ,
       indexrelid ,
       schemaname ,
       relname ,
       indexrelname ,
       idx_blks_read ,
       idx_blks_hit
FROM pg_statio_all_indexes;


INSERT INTO db_stat.snapshot_pg_statio_all_tables(relid ,schemaname ,relname ,heap_blks_read ,heap_blks_hit ,idx_blks_read ,idx_blks_hit ,toast_blks_read ,toast_blks_hit ,tidx_blks_read ,tidx_blks_hit)
SELECT relid ,
       schemaname ,
       relname ,
       heap_blks_read ,
       heap_blks_hit ,
       idx_blks_read ,
       idx_blks_hit ,
       toast_blks_read ,
       toast_blks_hit ,
       tidx_blks_read ,
       tidx_blks_hit
FROM pg_statio_all_tables;

--
select db_stat.process_snapshot_table('192.168.174.10',5432,'MASTER','');

--创建一个shell脚本,每天通过定时任务进行信息采集
cat snap_stat.sh

#!/bin/sh
source ~/.bash_profile
source /etc/profile
PSQL="psql"


help_msg (){
        echo ""
        echo "Usage:"
        echo "  -f              要输出结果的文件,如果为null,则默认为/tmp/snapshot_pg_stat.log"
        echo "  -u              数据库连接用户名,如果为null,则为postgresql默认"
        echo "  -d              连接的数据库名,如果为null,则为postgresql默认"
        echo "  -H              数据库的主机ip,如果为null,则为postgresql默认"
        echo "  -p              数据库的端口,如果为null,则为postgresql默认"
        echo "  -m              数据库的类型,MASTER为主,SLAVE为从"
        echo ""
        exit 0
}

# end functions

while getopts "f:u:d:H:p:m:" flag
do
        case $flag in
                f) FILENAME=$OPTARG
                        ;;
                u) USERNAME=$OPTARG
                        ;;
                d) DATABASE=$OPTARG
                        ;;
                H) HOST=$OPTARG
                        ;;
                p) PORT=$OPTARG
                        ;;
                m) DATABASE_TYPE=$OPTARG
                        ;;        
                \?|h) help_msg
                        ;;
        esac
done


if [ $USERNAME"x" == "x" ]
then
USERNAME=postgres
fi

if [ $DATABASE"x" == "x" ]
then
DATABASE=postgres
fi

if [ $HOST"x" == "x" ]
then
help_msg
fi

if [ $PORT"x" == "x" ]
then
PORT=5432
fi

if [ $DATABASE_TYPE"x" == "x" ]
then
DATABASE_TYPE=MASTER
fi

if [ $FILENAME"x" == "x" ]
then
FILENAME=/tmp/snapshot_pg_stat.log
fi

OUTPUT_FILENAME=/tmp/snapshot_pg_stat.csv

echo "" > $FILENAME


if [ ! -f $FILENAME ]
        then
        touch $FILENAME
else
    printf "" | tee -a $FILENAME
fi


echo "脚本于时间 `date "+%Y-%m-%d %H:%M:%S"` 开始执行" >> $FILENAME

echo "脚本开始于`date "+%Y-%m-%d %H:%M:%S"` 处理pg_stat_all_indexes表" >> $FILENAME
$PSQL -p $PORT -U $USERNAME -d $DATABASE -c "copy (select relid ,indexrelid ,schemaname ,relname ,indexrelname,idx_scan ,idx_tup_read,idx_tup_fetch from pg_stat_all_indexes) to '$OUTPUT_FILENAME' with csv"
$PSQL -p 5432 -U postgres -d postgres -h 192.168.174.11 -c "\copy db_stat.snapshot_pg_stat_all_indexes(relid ,indexrelid ,schemaname ,relname ,indexrelname,idx_scan ,idx_tup_read,idx_tup_fetch) from '$OUTPUT_FILENAME' with csv"
echo "脚本开始于`date "+%Y-%m-%d %H:%M:%S"` 处理pg_stat_all_tables表" >> $FILENAME
$PSQL -p $PORT -U $USERNAME -d $DATABASE -c "copy (select relid ,schemaname ,relname ,seq_scan ,seq_tup_read ,idx_scan ,idx_tup_fetch ,n_tup_ins ,n_tup_upd ,n_tup_del ,n_tup_hot_upd ,n_live_tup ,n_dead_tup ,last_vacuum ,last_autovacuum ,last_analyze ,last_autoanalyze ,vacuum_count ,autovacuum_count ,analyze_count ,autoanalyze_count from pg_stat_all_tables) to '$OUTPUT_FILENAME' with csv"
$PSQL -p 5432 -U postgres -d postgres -h 192.168.174.11 -c "\copy db_stat.snapshot_pg_stat_all_tables(relid ,schemaname ,relname ,seq_scan ,seq_tup_read ,idx_scan ,idx_tup_fetch ,n_tup_ins ,n_tup_upd ,n_tup_del ,n_tup_hot_upd ,n_live_tup ,n_dead_tup ,last_vacuum ,last_autovacuum ,last_analyze ,last_autoanalyze ,vacuum_count ,autovacuum_count ,analyze_count ,autoanalyze_count) from '$OUTPUT_FILENAME' with csv"
echo "脚本开始于`date "+%Y-%m-%d %H:%M:%S"` 处理pg_statio_all_indexes表" >> $FILENAME
$PSQL -p $PORT -U $USERNAME -d $DATABASE -c "copy (select relid ,indexrelid ,schemaname ,relname ,indexrelname ,idx_blks_read ,idx_blks_hit from pg_statio_all_indexes) to '$OUTPUT_FILENAME' with csv"
$PSQL -p 5432 -U postgres -d postgres -h 192.168.174.11 -c "\copy db_stat.snapshot_pg_statio_all_indexes(relid ,indexrelid ,schemaname ,relname ,indexrelname ,idx_blks_read ,idx_blks_hit) from '$OUTPUT_FILENAME' with csv"
echo "脚本开始于`date "+%Y-%m-%d %H:%M:%S"` 处理pg_statio_all_tables表" >> $FILENAME
$PSQL -p $PORT -U $USERNAME -d $DATABASE -c "copy (select relid ,schemaname ,relname ,heap_blks_read ,heap_blks_hit ,idx_blks_read ,idx_blks_hit ,toast_blks_read ,toast_blks_hit ,tidx_blks_read ,tidx_blks_hit from pg_statio_all_tables) to '$OUTPUT_FILENAME' with csv"
$PSQL -p 5432 -U postgres -d postgres -h 192.168.174.11 -c "\copy db_stat.snapshot_pg_statio_all_tables(relid ,schemaname ,relname ,heap_blks_read ,heap_blks_hit ,idx_blks_read ,idx_blks_hit ,toast_blks_read ,toast_blks_hit ,tidx_blks_read ,tidx_blks_hit) from '$OUTPUT_FILENAME' with csv"

$PSQL -p 5432 -U postgres -d postgres -h 192.168.174.11 -c "select db_stat.process_snapshot_table('$HOST',$PORT,'$DATABASE_TYPE','database stat snapshot');"

echo "############################################################################################" >> $FILENAME
echo "脚本于时间 `date "+%Y-%m-%d %H:%M:%S"` 结束执行" >> $FILENAME




--清空数据表
truncate table db_stat.snapshot_pg_stat_all_indexes ;
truncate table db_stat.snapshot_pg_stat_all_tables ;
truncate table db_stat.snapshot_pg_statio_all_indexes ;
truncate table db_stat.snapshot_pg_statio_all_tables ;

--手动执行shell脚本
./snap_stat.sh -d mydb -p 5432 -m SLAVE -u postgres -H 192.168.174.10
--定时任务,每天8点开始执行
8 8 * * * /db/pgsql/snap_stat.sh -d mydb -p 5435 -m SLAVE -u postgres -H 192.168.174.10



--查看使用比较少的索引
select * 
      from (
           SELECT t.relname,
                  t.indexrelname ,
                  max(idx_scan)-min(idx_scan) AS diff_idx_scan,
                  max(idx_tup_read)-min(idx_tup_read) AS diff_idx_tup_read
           FROM db_stat.snapshot_pg_stat_all_indexes t
           --WHERE snap_create_time BETWEEN '2015-12-11' AND '2016-03-11'
           GROUP BY t.relname, t.indexrelname) t1
order by diff_idx_scan,relname,indexrelname ;

--查看索引使用率趋势图
select relname,
       indexrelname,
       snap_day,
       diff_idx_scan,
       case when sum(diff_idx_scan) over w1 >0 then  diff_idx_scan*100/sum(diff_idx_scan) over w1 else 0 end as  diff_idx_scan_percent,
       diff_idx_tup_read,
       case when sum(diff_idx_tup_read) over w1 >0 then  diff_idx_tup_read*100/sum(diff_idx_tup_read) over w1  else 0 end as diff_idx_tup_read_percent 
from (
      SELECT t.relname,
             t.indexrelname,
             date_trunc('hour', snap_create_time) snap_day,
             t.idx_scan-lag(t.idx_scan,1) over w AS diff_idx_scan,
             t.idx_tup_read - lag(t.idx_tup_read,1) over w AS diff_idx_tup_read
      from db_stat.snapshot_pg_stat_all_indexes t 
      --where indexrelname in ('','')
      WINDOW w AS (PARTITION BY t.relname,t.indexrelname ORDER BY date_trunc('hour', t.snap_create_time))
) t1 
where diff_idx_scan is not null
WINDOW w1 as (PARTITION BY t1.relname,t1.indexrelname)
order by relname,indexrelname,snap_day;

相关实践学习
使用PolarDB和ECS搭建门户网站
本场景主要介绍基于PolarDB和ECS实现搭建门户网站。
阿里云数据库产品家族及特性
阿里云智能数据库产品团队一直致力于不断健全产品体系,提升产品性能,打磨产品功能,从而帮助客户实现更加极致的弹性能力、具备更强的扩展能力、并利用云设施进一步降低企业成本。以云原生+分布式为核心技术抓手,打造以自研的在线事务型(OLTP)数据库Polar DB和在线分析型(OLAP)数据库Analytic DB为代表的新一代企业级云原生数据库产品体系, 结合NoSQL数据库、数据库生态工具、云原生智能化数据库管控平台,为阿里巴巴经济体以及各个行业的企业客户和开发者提供从公共云到混合云再到私有云的完整解决方案,提供基于云基础设施进行数据从处理、到存储、再到计算与分析的一体化解决方案。本节课带你了解阿里云数据库产品家族及特性。
目录
相关文章
|
5月前
|
存储 关系型数据库 分布式数据库
PolarDB产品使用问题之如何查看PolarDB for PostgreSQL的备份信息
PolarDB产品使用合集涵盖了从创建与管理、数据管理、性能优化与诊断、安全与合规到生态与集成、运维与支持等全方位的功能和服务,旨在帮助企业轻松构建高可用、高性能且易于管理的数据库环境,满足不同业务场景的需求。用户可以通过阿里云控制台、API、SDK等方式便捷地使用这些功能,实现数据库的高效运维与持续优化。
|
7月前
|
SQL 关系型数据库 PostgreSQL
PostgreSQL【SQL 01】根据条件更新字段值或追加信息STRPOS(string, substring)函数使用及LIKE函数对比
PostgreSQL【SQL 01】根据条件更新字段值或追加信息STRPOS(string, substring)函数使用及LIKE函数对比
168 0
|
SQL 关系型数据库 Java
PostgreSQL统计信息的几个重要视图
PostgreSQL统计信息的几个重要视图
285 1
|
存储 SQL 关系型数据库
PostgreSQL插件HypoPG:支持虚拟索引
PostgreSQL插件HypoPG:支持虚拟索引
399 0
|
存储 缓存 关系型数据库
PostgreSQL 14新特性--减少索引膨胀
PostgreSQL 14新特性--减少索引膨胀
482 0
|
关系型数据库 PostgreSQL 索引
PostgreSQL通过索引获取heap tuple解析
PostgreSQL通过索引获取heap tuple解析
168 0
|
存储 SQL 关系型数据库
PostgreSQL的B-tree索引(下)
PostgreSQL的B-tree索引(下)
124 0
|
存储 算法 关系型数据库
PostgreSQL的B-tree索引(上)
PostgreSQL的B-tree索引
130 0
|
存储 SQL 人工智能
4 PostgreSQL 索引,全文检索,模糊匹配,近似度匹配(三)
4 PostgreSQL 索引,全文检索,模糊匹配,近似度匹配(三)
4 PostgreSQL 索引,全文检索,模糊匹配,近似度匹配(三)
|
Cloud Native 关系型数据库 MySQL
直播预告 | PostgreSQL 内核解读系列第六讲:PostgreSQL 索引介绍(下)
本系列课程将面向DBA、高校学生、内核爱好者,分15个章节,系统化介绍PG核心技术原理、用法和代码实现。希望通过课程学习,让没有内核经验的同学和DBA,也可以进行简单的特性开发,更深入理解PG配置和运行原理。本节课将讲述 PostgreSQL 索引基本介绍,涵盖Hash索引、Bitmap索引、GiST索引、GIN索引、brin索引、其他索引等内容。
直播预告 | PostgreSQL 内核解读系列第六讲:PostgreSQL 索引介绍(下)