postgresql 定时收集表和索引统计信息

--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;

上一篇:安装hadoop集群(Multi Cluster)


下一篇:如何利用多核CPU来加速你的Linux命令