学习目标
学习openGauss收集统计信息、打印执行计划、垃圾收集和checkpoint
课程学习
连接数据库
#第一次进入等待15秒
#数据库启动中...
su -
omm
gsql
-r
1.准备数据
Create
schema tpcds;
CREATE
TABLE tpcds.customer_address
(
ca_address_sk
integer NOT NULL ,
ca_address_id
character(16),
ca_street_number
character(10) ,
ca_street_name
character varying(60) ,
ca_street_type
character(15) ,
ca_suite_number
character(10) ,
ca_city
character varying(60) ,
ca_county
character varying(30) ,
ca_state
character(2) ,
ca_zip
character(10) ,
ca_country
character varying(20) ,
ca_gmt_offset
numeric(5,2) ,
ca_location_type
character(20)
);
insert
into tpcds.customer_address values
(1,
'AAAAAAAABAAAAAAA', '18', 'Jackson', 'Parkway', 'Suite 280', 'Fairfield',
'Maricopa County', 'AZ', '86192' ,'United States', -7.00, 'condo'),
(2,
'AAAAAAAACAAAAAAA', '362', 'Washington 6th', 'RD', 'Suite 80', 'Fairview',
'Taos County', 'NM', '85709', 'United States', -7.00, 'condo'),
(3,
'AAAAAAAADAAAAAAA', '585', 'Dogwood Washington', 'Circle', 'Suite Q', 'Pleasant
Valley', 'York County', 'PA', '12477', 'United States', -5.00, 'single
family');
–使用序列的generate_series(1,N)函数对表插入数据
insert
into tpcds.customer_address values(generate_series(10, 10000));
2.收集统计信息
–查看系统表中表的统计信息
select
relname, relpages, reltuples from pg_class where relname = 'customer_address';
—使用ANALYZE VERBOSE语句更新统计信息，并输出表的相关信息
analyze
VERBOSE tpcds.customer_address;
–查看系统表中表的统计信息
select
relname, relpages, reltuples from pg_class where relname = 'customer_address';
3.打印执行计划
–使用默认的打印格式
SET
explain_perf_mode=normal;
–显示表简单查询的执行计划
EXPLAIN
SELECT * FROM tpcds.customer_address;
–以JSON格式输出的执行计划（explain_perf_mode为normal时）
EXPLAIN(FORMAT
JSON) SELECT * FROM tpcds.customer_address;
–禁止开销估计的执行计划
EXPLAIN(COSTS
FALSE)SELECT * FROM tpcds.customer_address;
–带有聚集函数查询的执行计划
EXPLAIN
SELECT SUM(ca_address_sk) FROM tpcds.customer_address WHERE
ca_address_sk<100;
–有索引条件的执行计划
create
index customer_address_idx on tpcds.customer_address(ca_address_sk);
EXPLAIN
SELECT * FROM tpcds.customer_address WHERE ca_address_sk<100;
4.垃圾收集
–VACUUM回收表或B-Tree索引中已经删除的行所占据的存储空间
update
tpcds.customer_address set ca_address_sk = ca_address_sk + 1 where
ca_address_sk <100;
VACUUM
(VERBOSE, ANALYZE) tpcds.customer_address;
5.事务日志检查点
–检查点（CHECKPOINT）是一个事务日志中的点，所有数据文件都在该点被更新以反映日志中的信息，所有数据文件都将被刷新到磁盘
CHECKPOINT;
6.清理数据
drop
schema tpcds cascade;
课后作业
1. 创建分区表，并用generate_series(1,N)函数对表插入数据
omm=# create table
students(s_id int ,s_name varchar(30))
omm-# partition by
range(s_id)
omm-# (
omm(# partition p1 values less than(10),
omm(# partition p2 values
less than(20),
omm(# partition p3
values less than(10000));
CREATE TABLE
omm=# insert into
students values(generate_series(4,2000));
omm=# INSERT 0 1997
2. 收集表统计信息
omm=# select
relname, relpages, reltuples from pg_class where relname = 'students';
relname
| relpages | reltuples
----------+----------+-----------
students | 0 | 0
(1 row)
3. 显示简单查询的执行计划；建立索引并显示有索引条件的执行计划
omm=# EXPLAIN
SELECT * FROM students;
QUERY
PLAN
------------------------------------------------------------------------------
Partition Iterator (cost=0.00..19.02 rows=802 width=82)
Iterations: 3
->
Partitioned Seq Scan on students
(cost=0.00..19.02 rows=802 width=82)
Selected Partitions: 1..3
(4 rows)
omm=# create index
student_index on students(s_id);
CREATE INDEX
omm=# EXPLAIN
SELECT * FROM students where s_id<100;
QUERY PLAN
------------------------------------------------------------------------------
Partition Iterator (cost=0.00..21.02 rows=267 width=82)
Iterations: 3
->
Partitioned Seq Scan on students
(cost=0.00..21.02 rows=267 width=82)
Filter: (s_id < 100)
Selected Partitions: 1..3
(5 rows)
4.更新表数据，并做垃圾收集
omm=# update
students set s_id = s_id+100;
UPDATE 1997
omm=# VACUUM
(VERBOSE, ANALYZE)students ;
5.清理数据
omm=# drop table students cascade;
DROP TABLE
