KingbaseES例程之拥有大量索引的表导入数据
概述
如何快速插入大量数据比如几千万上亿的带索引的数据表。
数据准备
准备一个拥有二十个索引的数据表。
kingbase=# \d+ bigtab
Table "kingbase.bigtab"
Column | Type | Collation | Nullable | Default | Storage | Stats target | Description
--------+---------+-----------+----------+---------+----------+--------------+-------------
id | integer | | | | plain | |
c01 | integer | | | | plain | |
c02 | integer | | | | plain | |
c03 | integer | | | | plain | |
c04 | integer | | | | plain | |
c05 | integer | | | | plain | |
c06 | integer | | | | plain | |
c07 | integer | | | | plain | |
c08 | integer | | | | plain | |
c09 | integer | | | | plain | |
c10 | integer | | | | plain | |
c11 | integer | | | | plain | |
c12 | integer | | | | plain | |
c13 | integer | | | | plain | |
c14 | integer | | | | plain | |
c15 | integer | | | | plain | |
c16 | integer | | | | plain | |
c17 | integer | | | | plain | |
c18 | integer | | | | plain | |
c19 | integer | | | | plain | |
c20 | integer | | | | plain | |
c21 | integer | | | | plain | |
c22 | integer | | | | plain | |
c23 | integer | | | | plain | |
c24 | integer | | | | plain | |
c25 | integer | | | | plain | |
c26 | integer | | | | plain | |
c27 | integer | | | | plain | |
c28 | integer | | | | plain | |
c29 | integer | | | | plain | |
t01 | text | | | | extended | |
t02 | text | | | | extended | |
t03 | text | | | | extended | |
t04 | text | | | | extended | |
t05 | text | | | | extended | |
t06 | text | | | | extended | |
t07 | text | | | | extended | |
t08 | text | | | | extended | |
t09 | text | | | | extended | |
t10 | text | | | | extended | |
t11 | text | | | | extended | |
t12 | text | | | | extended | |
t13 | text | | | | extended | |
t14 | text | | | | extended | |
t15 | text | | | | extended | |
t16 | text | | | | extended | |
t17 | text | | | | extended | |
t18 | text | | | | extended | |
t19 | text | | | | extended | |
t20 | text | | | | extended | |
Indexes:
"bigtab_i01" btree (c01)
"bigtab_i02" btree (c02)
"bigtab_i03" btree (c03)
"bigtab_i04" btree (c04)
"bigtab_i05" btree (c05)
"bigtab_i06" btree (c06)
"bigtab_i07" btree (c07)
"bigtab_i08" btree (c08)
"bigtab_i09" btree (c09)
"bigtab_i10" btree (c10)
"bigtab_i11" btree (c11)
"bigtab_i12" btree (c12)
"bigtab_i13" btree (c13)
"bigtab_i14" btree (c14)
"bigtab_i15" btree (c15)
"bigtab_i16" btree (c16)
"bigtab_i17" btree (c17)
"bigtab_i18" btree (c18)
"bigtab_i19" btree (c19)
"bigtab_i20" btree (c20)
Access method: heap
kingbase=#
方法一：直接插入海量数据，自动维护索引
kingbase=#
kingbase=# insert into bigtab
kingbase-# select id
kingbase-# , (random() * 100)::int + 1000 c01
kingbase-# , (random() * 200)::int + 1000 c02
kingbase-# , (random() * 300)::int + 10000 c03
kingbase-# , (random() * 400)::int + 10000 c04
kingbase-# , (random() * 500)::int + 10000 c05
kingbase-# , (random() * 600)::int + 10000 c06
kingbase-# , (random() * 700)::int + 10000 c07
kingbase-# , (random() * 800)::int + 10000 c08
kingbase-# , (random() * 900)::int + 10000 c09
kingbase-# , (random() * 1000)::int + 10000 c10
kingbase-# , (random() * 2000)::int + 10000 c11
kingbase-# , (random() * 3000)::int + 10000 c12
kingbase-# , (random() * 4000)::int + 10000 c13
kingbase-# , (random() * 5000)::int + 10000 c14
kingbase-# , (random() * 6000)::int + 10000 c15
kingbase-# , (random() * 7000)::int + 10000 c16
kingbase-# , (random() * 8000)::int + 10000 c17
kingbase-# , (random() * 9000)::int + 10000 c18
kingbase-# , (random() * 10000)::int + 10000 c19
kingbase-# , (random() * 20000)::int + 10000 c20
kingbase-# , (random() * 30000)::int + 10000 c21
kingbase-# , (random() * 40000)::int + 10000 c22
kingbase-# , (random() * 50000)::int + 10000 c23
kingbase-# , (random() * 60000)::int + 10000 c24
kingbase-# , (random() * 70000)::int + 10000 c25
kingbase-# , (random() * 80000)::int + 10000 c26
kingbase-# , (random() * 90000)::int + 10000 c27
kingbase-# , (random() * 10000)::int + 10000 c28
kingbase-# , (random() * 10000)::int + 10000 c29
kingbase-# , md5(random()::text) t01
kingbase-# , md5(random()::text) t02
kingbase-# , md5(random()::text) t03
kingbase-# , md5(random()::text) t04
kingbase-# , md5(random()::text) t05
kingbase-# , md5(random()::text) t06
kingbase-# , md5(random()::text) t07
kingbase-# , md5(random()::text) t08
kingbase-# , md5(random()::text) t09
kingbase-# , md5(random()::text) t10
kingbase-# , md5(random()::text) t11
kingbase-# , md5(random()::text) t12
kingbase-# , md5(random()::text) t13
kingbase-# , md5(random()::text) t14
kingbase-# , md5(random()::text) t15
kingbase-# , md5(random()::text) t16
kingbase-# , md5(random()::text) t17
kingbase-# , md5(random()::text) t18
kingbase-# , md5(random()::text) t19
kingbase-# , md5(random()::text) t20
kingbase-# from generate_series(1, 2000000) id;
INSERT 0 2000000
Time: 299331.143 ms (04:59.331)
优点： 语句单一；自动维护索引；自动支持之后的索引。
缺点： 逐行维护索引，造成用时较长。
方法二：删除索引，插入海量数据，再创建索引
kingbase=#
kingbase=# do
kingbase-# $$
kingbase$# begin
kingbase$# drop index bigtab_i01;
kingbase$# drop index bigtab_i02;
kingbase$# drop index bigtab_i03;
kingbase$# drop index bigtab_i04;
kingbase$# drop index bigtab_i05;
kingbase$# drop index bigtab_i06;
kingbase$# drop index bigtab_i07;
kingbase$# drop index bigtab_i08;
kingbase$# drop index bigtab_i09;
kingbase$# drop index bigtab_i10;
kingbase$# drop index bigtab_i11;
kingbase$# drop index bigtab_i12;
kingbase$# drop index bigtab_i13;
kingbase$# drop index bigtab_i14;
kingbase$# drop index bigtab_i15;
kingbase$# drop index bigtab_i16;
kingbase$# drop index bigtab_i17;
kingbase$# drop index bigtab_i18;
kingbase$# drop index bigtab_i19;
kingbase$# drop index bigtab_i20;
kingbase$#
kingbase$# insert into bigtab
kingbase$# select id
kingbase$# , (random() * 100)::int + 1000 c01
kingbase$# , (random() * 200)::int + 1000 c02
kingbase$# , (random() * 300)::int + 10000 c03
kingbase$# , (random() * 400)::int + 10000 c04
kingbase$# , (random() * 500)::int + 10000 c05
kingbase$# , (random() * 600)::int + 10000 c06
kingbase$# , (random() * 700)::int + 10000 c07
kingbase$# , (random() * 800)::int + 10000 c08
kingbase$# , (random() * 900)::int + 10000 c09
kingbase$# , (random() * 1000)::int + 10000 c10
kingbase$# , (random() * 2000)::int + 10000 c11
kingbase$# , (random() * 3000)::int + 10000 c12
kingbase$# , (random() * 4000)::int + 10000 c13
kingbase$# , (random() * 5000)::int + 10000 c14
kingbase$# , (random() * 6000)::int + 10000 c15
kingbase$# , (random() * 7000)::int + 10000 c16
kingbase$# , (random() * 8000)::int + 10000 c17
kingbase$# , (random() * 9000)::int + 10000 c18
kingbase$# , (random() * 10000)::int + 10000 c19
kingbase$# , (random() * 20000)::int + 10000 c20
kingbase$# , (random() * 30000)::int + 10000 c21
kingbase$# , (random() * 40000)::int + 10000 c22
kingbase$# , (random() * 50000)::int + 10000 c23
kingbase$# , (random() * 60000)::int + 10000 c24
kingbase$# , (random() * 70000)::int + 10000 c25
kingbase$# , (random() * 80000)::int + 10000 c26
kingbase$# , (random() * 90000)::int + 10000 c27
kingbase$# , (random() * 10000)::int + 10000 c28
kingbase$# , (random() * 10000)::int + 10000 c29
kingbase$# , md5(random()::text) t01
kingbase$# , md5(random()::text) t02
kingbase$# , md5(random()::text) t03
kingbase$# , md5(random()::text) t04
kingbase$# , md5(random()::text) t05
kingbase$# , md5(random()::text) t06
kingbase$# , md5(random()::text) t07
kingbase$# , md5(random()::text) t08
kingbase$# , md5(random()::text) t09
kingbase$# , md5(random()::text) t10
kingbase$# , md5(random()::text) t11
kingbase$# , md5(random()::text) t12
kingbase$# , md5(random()::text) t13
kingbase$# , md5(random()::text) t14
kingbase$# , md5(random()::text) t15
kingbase$# , md5(random()::text) t16
kingbase$# , md5(random()::text) t17
kingbase$# , md5(random()::text) t18
kingbase$# , md5(random()::text) t19
kingbase$# , md5(random()::text) t20
kingbase$# from generate_series(1, 2000000) id;
kingbase$#
kingbase$# create index bigtab_i01 on bigtab (c01);
kingbase$# create index bigtab_i02 on bigtab (c02);
kingbase$# create index bigtab_i03 on bigtab (c03);
kingbase$# create index bigtab_i04 on bigtab (c04);
kingbase$# create index bigtab_i05 on bigtab (c05);
kingbase$# create index bigtab_i06 on bigtab (c06);
kingbase$# create index bigtab_i07 on bigtab (c07);
kingbase$# create index bigtab_i08 on bigtab (c08);
kingbase$# create index bigtab_i09 on bigtab (c09);
kingbase$# create index bigtab_i10 on bigtab (c10);
kingbase$# create index bigtab_i11 on bigtab (c11);
kingbase$# create index bigtab_i12 on bigtab (c12);
kingbase$# create index bigtab_i13 on bigtab (c13);
kingbase$# create index bigtab_i14 on bigtab (c14);
kingbase$# create index bigtab_i15 on bigtab (c15);
kingbase$# create index bigtab_i16 on bigtab (c16);
kingbase$# create index bigtab_i17 on bigtab (c17);
kingbase$# create index bigtab_i18 on bigtab (c18);
kingbase$# create index bigtab_i19 on bigtab (c19);
kingbase$# create index bigtab_i20 on bigtab (c20);
kingbase$#
kingbase$# end;
kingbase$# $$;
ANONYMOUS BLOCK
Time: 83069.170 ms (01:23.069)
优点： 批量维护索引，用时最短。
缺点： 语句复杂且固化；手动维护删建索引语句；不支持之后的索引。
方法三：禁止索引更改，插入海量数据，重建表的全部索引
kingbase=# do
kingbase-# $$
kingbase$# begin
kingbase$#
kingbase$# update pg_index
kingbase$# set indislive= false
kingbase$# where indrelid = 'bigtab'::regclass;
kingbase$#
kingbase$# insert into bigtab
kingbase$# select id
kingbase$# , (random() * 100)::int + 1000 c01
kingbase$# , (random() * 200)::int + 1000 c02
kingbase$# , (random() * 300)::int + 10000 c03
kingbase$# , (random() * 400)::int + 10000 c04
kingbase$# , (random() * 500)::int + 10000 c05
kingbase$# , (random() * 600)::int + 10000 c06
kingbase$# , (random() * 700)::int + 10000 c07
kingbase$# , (random() * 800)::int + 10000 c08
kingbase$# , (random() * 900)::int + 10000 c09
kingbase$# , (random() * 1000)::int + 10000 c10
kingbase$# , (random() * 2000)::int + 10000 c11
kingbase$# , (random() * 3000)::int + 10000 c12
kingbase$# , (random() * 4000)::int + 10000 c13
kingbase$# , (random() * 5000)::int + 10000 c14
kingbase$# , (random() * 6000)::int + 10000 c15
kingbase$# , (random() * 7000)::int + 10000 c16
kingbase$# , (random() * 8000)::int + 10000 c17
kingbase$# , (random() * 9000)::int + 10000 c18
kingbase$# , (random() * 10000)::int + 10000 c19
kingbase$# , (random() * 20000)::int + 10000 c20
kingbase$# , (random() * 30000)::int + 10000 c21
kingbase$# , (random() * 40000)::int + 10000 c22
kingbase$# , (random() * 50000)::int + 10000 c23
kingbase$# , (random() * 60000)::int + 10000 c24
kingbase$# , (random() * 70000)::int + 10000 c25
kingbase$# , (random() * 80000)::int + 10000 c26
kingbase$# , (random() * 90000)::int + 10000 c27
kingbase$# , (random() * 10000)::int + 10000 c28
kingbase$# , (random() * 10000)::int + 10000 c29
kingbase$# , md5(random()::text) t01
kingbase$# , md5(random()::text) t02
kingbase$# , md5(random()::text) t03
kingbase$# , md5(random()::text) t04
kingbase$# , md5(random()::text) t05
kingbase$# , md5(random()::text) t06
kingbase$# , md5(random()::text) t07
kingbase$# , md5(random()::text) t08
kingbase$# , md5(random()::text) t09
kingbase$# , md5(random()::text) t10
kingbase$# , md5(random()::text) t11
kingbase$# , md5(random()::text) t12
kingbase$# , md5(random()::text) t13
kingbase$# , md5(random()::text) t14
kingbase$# , md5(random()::text) t15
kingbase$# , md5(random()::text) t16
kingbase$# , md5(random()::text) t17
kingbase$# , md5(random()::text) t18
kingbase$# , md5(random()::text) t19
kingbase$# , md5(random()::text) t20
kingbase$# from generate_series(1, 2000000) id;
kingbase$#
kingbase$# update pg_index
kingbase$# set indislive= true
kingbase$# where indrelid = 'bigtab'::regclass;
kingbase$#
kingbase$# analyse bigtab;
kingbase$# reindex table bigtab;
kingbase$#
kingbase$# end;
kingbase$# $$;
ANONYMOUS BLOCK
Time: 87110.126 ms (01:27.110)
优点： 批量维护索引，用时短；语句固定模式；自动维护索引；支持之后的索引。
缺点： 多个SQL语句，不易嵌入语句块。
最后的话
reindex table 的执行依赖统计信息，所以需要执行 analyse table ，才能成功重建表的全部可更新的索引。
reindex index 不受上述因素的影响，可以强制重建不更新的索引，并自动修改 indislive= true。
如果在REINDEX期间出现异常，那么所有需要rebuild的索引的状态都是invalid，意味着这些索引仍然占用空间，定义仍在但不能使用。
避免REINDEX期间出现异常，可以在索引更新操作时，跳过唯一索引和外键依赖索引等。
KINGBASE研究院
