openGauss每日一练第19天 | 收集统计信息，打印执行计划

原创 Sean 2021-12-19

315

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

课程学习实操

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=# Create schema tpcds;
CREATE SCHEMA
omm=# CREATE TABLE tpcds.customer_address
omm-# (
omm(# ca_address_sk integer NOT NULL ,
omm(# ca_address_id character(16),
omm(# ca_street_number character(10) ,
omm(# ca_street_name character varying(60) ,
omm(# ca_street_type character(15) ,
omm(# ca_suite_number character(10) ,
omm(# ca_city character varying(60) ,
omm(# ca_county character varying(30) ,
omm(# ca_state character(2) ,
omm(# ca_zip character(10) ,
omm(# ca_country character varying(20) ,
omm(# ca_gmt_offset numeric(5,2) ,
omm(# ca_location_type character(20)
omm(# );
CREATE TABLE
omm=# insert into tpcds.customer_address values
omm-# (1, 'AAAAAAAABAAAAAAA', '18', 'Jackson', 'Parkway', 'Suite 280', 'Fairfield', 'Maricopa County', 'AZ', '86192' ,'United States', -7.00, 'condo'),
omm-# (2, 'AAAAAAAACAAAAAAA', '362', 'Washington 6th', 'RD', 'Suite 80', 'Fairview', 'Taos County', 'NM', '85709', 'United States', -7.00, 'condo'),
omm-# (3, 'AAAAAAAADAAAAAAA', '585', 'Dogwood Washington', 'Circle', 'Suite Q', 'Pleasant Valley', 'York County', 'PA', '12477', 'United States', -5.00, 'single family');
INSERT 0 3
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   
---------------+------------------+------------------+--------------------+-----------------+-----------------+-------
----------+-----------------+----------+------------+---------------+---------------+----------------------
             1 | AAAAAAAABAAAAAAA | 18               | Jackson            | Parkway         | Suite 280       | Fairfi
eld       | Maricopa County | AZ       | 86192      | United States |         -7.00 | condo               
             2 | AAAAAAAACAAAAAAA | 362              | Washington 6th     | RD              | Suite 80        | Fairvi
             3 | AAAAAAAADAAAAAAA | 585              | Dogwood Washington | Circle          | Suite Q         | Pleasa
nt Valley | York County     | PA       | 12477      | United States |         -5.00 | single family       
(3 rows)

ew        | Taos County     | NM       | 85709      | United States |         -7.00 | condo               
omm=# 


复制

–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=# insert into tpcds.customer_address values(generate_series(10, 10000));
INSERT 0 9991
omm=# 
复制

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)

omm=# 
复制

—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;

omm=# analyze VERBOSE tpcds.customer_address;
INFO:  analyzing "tpcds.customer_address"(gaussdb pid=1)
INFO:  ANALYZE INFO : "customer_address": scanned 55 of 55 pages, containing 9994 live rows and 0 dead rows; 9994 rows in sample, 9994 estimated total rows(gaussdb pid=1)
ANALYZE
omm=# 
复制

–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)

omm=# 
复制

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;

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)
复制

–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;
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                   +
     }                                     +
   }                                       +
 ]
(1 row)

复制

–禁止开销估计的执行计划


EXPLAIN(COSTS FALSE)SELECT * FROM tpcds.customer_address;

omm=# EXPLAIN(COSTS FALSE)SELECT * FROM tpcds.customer_address;
          QUERY PLAN          
------------------------------
 Seq Scan on customer_address
(1 row)

复制

–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;
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=# 
复制

–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=# create index customer_address_idx on tpcds.customer_address(ca_address_sk);
CREATE INDEX
omm=# EXPLAIN SELECT * FROM tpcds.customer_address WHERE ca_address_sk<100;
omm=#                                            QUERY PLAN                                           
------------------------------------------------------------------------------------------------
 [Bypass]
 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)


omm=# 
复制

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=# 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
omm=# 

复制

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;
omm=# CHECKPOINT;
CHECKPOINT
复制

6.清理数据

drop schema tpcds cascade;

omm=# drop schema tpcds cascade;
NOTICE:  drop cascades to table tpcds.customer_address
DROP SCHEMA
omm=# 
复制

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 partition_table
omm-# (
omm(#         c1 int,
omm(#         c2 CHAR(2)
omm(# )
omm-# partition by range (c1)
omm-# (
omm(#         partition partition_table_p0 values less than (10000),
omm(#         partition partition_table_p1 values less than (20000),
omm(#         partition partition_table_p2 values less than (30000)
omm(# );
CREATE TABLE
omm=# insert into partition_table values(generate_series(1, 29999));
INSERT 0 29999
omm=# 
复制

2.收集表统计信息

omm=# select relname, relpages, reltuples from pg_class where relname = 'partition_table';
     relname     | relpages | reltuples 
-----------------+----------+-----------
 partition_table |        0 |         0
(1 row)

omm=# analyze VERBOSE partition_table;
INFO:  analyzing "public.partition_table"(gaussdb pid=1)
INFO:  ANALYZE INFO : "partition_table": scanned 45 of 45 pages, containing 9999 live rows and 0 dead rows; 9999 rows in sample, 9999 estimated total rows(gaussdb pid=1)
INFO:  ANALYZE INFO : "partition_table": scanned 45 of 45 pages, containing 10000 live rows and 0 dead rows; 10000 rows in sample, 10000 estimated total rows(gaussdb pid=1)
INFO:  ANALYZE INFO : "partition_table": scanned 45 of 45 pages, containing 10000 live rows and 0 dead rows; 10000 rows in sample, 10000 estimated total rows(gaussdb pid=1)
ANALYZE
omm=# 
omm=# select relname, relpages, reltuples from pg_class where relname = 'partition_table';
     relname     | relpages | reltuples 
-----------------+----------+-----------
 partition_table |      135 |     29999
(1 row)

omm=# 
复制

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=# SET explain_perf_mode=normal;
SET
omm=# EXPLAIN SELECT * FROM partition_table;
                                       QUERY PLAN                                       
----------------------------------------------------------------------------------------
 Partition Iterator  (cost=0.00..434.99 rows=29999 width=16)
   Iterations: 3
   ->  Partitioned Seq Scan on partition_table  (cost=0.00..434.99 rows=29999 width=16)
         Selected Partitions:  1..3
(4 rows)

omm=# 

omm=# create index idx_c1 on partition_table(c1);
CREATE INDEX
omm=# explain select * from partition_table where c1=25;
                                  QUERY PLAN                                   
-------------------------------------------------------------------------------
 Index Scan using idx_c1 on partition_table  (cost=0.00..8.27 rows=1 width=16)
   Index Cond: (c1 = 25)
(2 rows)

omm=# 
复制

4.更新表数据，并做垃圾收集

omm=# update partition_table set c1 = c1 + 1 where c1 < 10000;
UPDATE 9999
omm=# VACUUM (VERBOSE, ANALYZE) partition_table;
INFO:  vacuuming "public.partition_table"(gaussdb pid=1)
INFO:  index "idx_c1" now contains 19997 row versions in 140 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:  "partition_table": found 0 removable, 19997 nonremovable row versions in 89 out of 89 pages(gaussdb pid=1)
DETAIL:  9999 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:  vacuuming "public.partition_table"(gaussdb pid=1)
INFO:  index "idx_c1" now contains 10001 row versions in 140 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:  "partition_table": found 0 removable, 10001 nonremovable row versions in 45 out of 45 pages(gaussdb pid=1)
DETAIL:  0 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:  vacuuming "public.partition_table"(gaussdb pid=1)
INFO:  index "idx_c1" now contains 10000 row versions in 140 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:  "partition_table": found 0 removable, 10000 nonremovable row versions in 45 out of 45 pages(gaussdb pid=1)
DETAIL:  0 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:  scanned index "idx_c1" to remove 0.000000 invisible rows(gaussdb pid=1)
DETAIL:  CPU 0.00s/0.00u sec elapsed 0.00 sec.
INFO:  analyzing "public.partition_table"(gaussdb pid=1)
INFO:  ANALYZE INFO : "partition_table": scanned 89 of 89 pages, containing 9998 live rows and 9999 dead rows; 9998 rows in sample, 9998 estimated total rows(gaussdb pid=1)
INFO:  ANALYZE INFO : "partition_table": scanned 45 of 45 pages, containing 10001 live rows and 0 dead rows; 7542 rows in sample, 10001 estimated total rows(gaussdb pid=1)
INFO:  ANALYZE INFO : "partition_table": scanned 45 of 45 pages, containing 10000 live rows and 0 dead rows; 7542 rows in sample, 10000 estimated total rows(gaussdb pid=1)
VACUUM
omm=# 
复制

5.清理数据

omm=# drop table partition_table;
DROP TABLE
omm=# 
复制

学习总结

通过本节课的学习，我掌握了统计信息的收集方法，SQL执行计划的查看方法，垃圾的回收的方法和检查点的使用。

opengauss

「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