openGauss每日一练第19天|《学习openGauss统计信息、执行计划、垃圾回收和checkpoint》学习心得体会和课后练习
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');
omm=# select * from tpcds.customer_address;
ca_address_sk | ca_address_id | ca_street_number | ca_street_name | ca_street_type |
ca_suite_number | ca_city | ca_county | ca_state | ca_zip | ca_country |
ca_gmt_offset | ca_location_type
---------------+------------------+------------------+--------------------+-----------------+-
-7.00 | condo
3 | AAAAAAAADAAAAAAA | 585 | Dogwood Washington | Circle |
----------------+-----------------+-----------------+----------+------------+---------------+-
--------------+----------------------
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 |
Suite Q | Pleasant Valley | York County | PA | 12477 | United States |
-5.00 | single family
(3 rows)
–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));
omm=# select ca_address_sk,ca_address_id from tpcds.customer_address limit 20;
ca_address_sk | ca_address_id
---------------+------------------
1 | AAAAAAAABAAAAAAA
2 | AAAAAAAACAAAAAAA
3 | AAAAAAAADAAAAAAA
10 |
11 |
12 |
13 |
14 |
15 |
16 |
17 |
18 |
19 |
20 |
21 |
22 |
23 |
24 |
25 |
26 |
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';
omm=# select relname, relpages, reltuples from pg_class where relname = 'customer_address';
relname | relpages | reltuples
------------------+----------+-----------
customer_address | 0 | 0
(1 row)
—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';
omm=# select relname, relpages, reltuples from pg_class where relname = 'customer_address';
relname | relpages | reltuples
------------------+----------+-----------
customer_address | 55 | 9994
(1 row)
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;
omm=# SET explain_perf_mode=normal;
SET
omm=# EXPLAIN SELECT * FROM tpcds.customer_address;
QUERY PLAN
-----------------------------------------------------------------------
Seq Scan on customer_address (cost=0.00..154.94 rows=9994 width=151)
(1 row)
omm=# EXPLAIN(FORMAT JSON) SELECT * FROM tpcds.customer_address;
QUERY PLAN
--------------------------------------------
[ +
{ +
"Plan": { +
"Node Type": "Seq Scan", +
"Relation Name": "customer_address",+
"Alias": "customer_address", +
"Startup Cost": 0.00, +
"Total Cost": 154.94, +
"Plan Rows": 9994, +
"Plan Width": 151 +
} +
omm=# } +
]
(1 row)
EXPLAIN(COSTS FALSE)SELECT * FROM tpcds.customer_address;
QUERY PLAN
------------------------------
Seq Scan on customer_address
(1 row)
omm=# EXPLAIN SELECT SUM(ca_address_sk) FROM tpcds.customer_address WHERE ca_address_sk<100;
QUERY PLAN
-------------------------------------------------------------------------
Aggregate (cost=180.16..180.17 rows=1 width=12)
-> Seq Scan on customer_address (cost=0.00..179.93 rows=94 width=4)
Filter: (ca_address_sk < 100)
(3 rows)
omm=# create index customer_address_idx on tpcds.customer_address(ca_address_sk);
EXPLAIN SELECT * FROM tpcds.customer_address WHERE ca_address_sk<100;CREATE INDEX
omm=#
Index Scan using customer_address_idx on customer_address (cost=0.00..9.90 rows=94 width=151
)
Index Cond: (ca_address_sk < 100)
(3 rows)
QUERY PLAN
----------------------------------------------------------------------------------------------
--
[Bypass]
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;
omm=# update tpcds.customer_address set ca_address_sk = ca_address_sk + 1 where ca_address_sk <100;
UPDATE 93
omm=#
omm=#
omm=# VACUUM (VERBOSE, ANALYZE) tpcds.customer_address;
INFO: vacuuming "tpcds.customer_address"(gaussdb pid=1)
INFO: index "customer_address_idx" now contains 10087 row versions in 31 pages(gaussdb pid=1)
DETAIL: 0 index row versions were removed.
0 index pages have been deleted, 0 are currently reusable.
CPU 0.00s/0.00u sec elapsed 0.00 sec.
INFO: "customer_address": found 0 removable, 10087 nonremovable row versions in 55 out of 55 pages(gaussdb pid=1)
DETAIL: 93 dead row versions cannot be removed yet.
There were 0 unused item pointers.
0 pages are entirely empty.
CPU 0.00s/0.00u sec elapsed 0.00 sec.
INFO: analyzing "tpcds.customer_address"(gaussdb pid=1)
INFO: ANALYZE INFO : "customer_address": scanned 55 of 55 pages, containing 9994 live rows and 93 dead rows; 9994 rows in sample, 9994 estimated total rows(gaussdb pid=1)
VACUUM
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;
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
create table test_table
(
c1 int,
c2 text
)
partition by range (c1)
(
partition test_table_p0 values less than (100),
partition test_table_p1 values less than (1000),
partition test_table_p2 values less than (15000)
);
INSERT INTO test_table SELECT id,md5(id::varchar) FROM generate_series(1,10000) AS id;
2.收集表统计信息
analyze VERBOSE test_table;
select relname, relpages, reltuples from pg_class where relname = 'test_table';
omm=#
analyze VERBOSE test_table;
INFO: analyzing "public.test_table"(gaussdb pid=1)
INFO: ANALYZE INFO : "test_table": scanned 1 of 1 pages, containing 99 live rows and 0 dead rows; 99 rows in sample, 99 estimated total rows(gaussdb pid=1)
INFO: ANALYZE INFO : "test_table": scanned 8 of 8 pages, containing 900 live rows and 0 dead rows; 900 rows in sample, 900 estimated total rows(gaussdb pid=1)
INFO: ANALYZE INFO : "test_table": scanned 76 of 76 pages, containing 9001 live rows and 0 dead rows; 9001 rows in sample, 9001 estimated total rows(gaussdb pid=1)
ANALYZE
omm=# select relname, relpages, reltuples from pg_class where relname = 'test_table';
relname | relpages | reltuples
------------+----------+-----------
test_table | 85 | 10000
(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
EXPLAIN SELECT * FROM test_table;
create index test_idx on test_table(c1);
EXPLAIN SELECT * FROM test_table WHERE c1<100;
omm=# EXPLAIN SELECT * FROM test_table WHERE c1<100;
QUERY PLAN
-------------------------------------------------------------------------------
Index Scan using test_idx on test_table (cost=0.00..10.00 rows=100 width=37)
Index Cond: (c1 < 100)
(2 rows)
4.更新表数据，并做垃圾收集
update test_table set c1 = c1 + 1 where c1 <1000;
VACUUM (VERBOSE, ANALYZE) test_table;
5.清理数据
drop table test_table;
「喜欢这篇文章，您的关注和赞赏是给作者最好的鼓励」
关注作者
【版权声明】本文为墨天轮用户原创内容，转载时必须标注文章的来源（墨天轮），文章链接，文章作者等基本信息，否则作者和墨天轮有权追究责任。如果您发现墨天轮中有涉嫌抄袭或者侵权的内容，欢迎发送邮件至：contact@modb.pro进行举报，并提供相关证据，一经查实，墨天轮将立刻删除相关内容。
