KingbaseES大数据量分区表添加主键与索引
KingbaseES大数据量分区表添加主键与索引
一、环境信息：
系统信息：
$ cat /etc/centos-release
CentOS Linux release 8.2.2004 (Core)
$ lscpu
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Byte Order: Little Endian
CPU(s): 1
On-line CPU(s) list: 0
Thread(s) per core: 1
Core(s) per socket: 1
Socket(s): 1
NUMA node(s): 1
Vendor ID: AuthenticAMD
CPU family: 23
Model: 24
Model name: AMD Ryzen 5 3500U with Radeon Vega Mobile Gfx
Stepping: 1
CPU MHz: 2096.060
BogoMIPS: 4192.12
Hypervisor vendor: KVM
Virtualization type: full
L1d cache: 32K
L1i cache: 64K
L2 cache: 512K
L3 cache: 4096K
NUMA node0 CPU(s): 0
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt rdtscp lm constant_tsc rep_good nopl nonstop_tsc cpuid extd_apicid tsc_known_freq pni pclmulqdq monitor ssse3 cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx rdrand hypervisor lahf_lm cr8_legacy abm sse4a misalignsse 3dnowprefetch ssbd vmmcall fsgsbase avx2 rdseed clflushopt arat
$ lsmem
RANGE SIZE STATE REMOVABLE BLOCK
0x0000000000000000-0x000000000fffffff 256M online no 0-1
0x0000000010000000-0x000000001fffffff 256M online yes 2-3
0x0000000020000000-0x0000000047ffffff 640M online no 4-8
0x0000000048000000-0x000000004fffffff 128M online yes 9
0x0000000050000000-0x000000007fffffff 768M online no 10-15
Memory block size: 128M
Total online memory: 2G
Total offline memory: 0B
$ df -Th
Filesystem Type Size Used Avail Use% Mounted on
devtmpfs devtmpfs 977M 0 977M 0% /dev
tmpfs tmpfs 994M 20K 994M 1% /dev/shm
tmpfs tmpfs 994M 8.5M 986M 1% /run
tmpfs tmpfs 994M 0 994M 0% /sys/fs/cgroup
/dev/mapper/cl_node2-root xfs 37G 2.9G 35G 8% /
/dev/sda1 ext4 976M 110M 799M 13% /boot
tmpfs tmpfs 199M 0 199M 0% /run/user/0
tmpfs tmpfs 199M 0 199M 0% /run/user/1000
二、初始化测试环境
# kingbase.conf除开启归档外，其余参数均为initdb初始化默认值
archive_mode=on
# 数据库版本信息
KingbaseES V008R006C007B0012 on x86_64-pc-linux-gnu, compiled by gcc (GCC) 4.1.2 20080704 (Red Hat 4.1.2-46), 64-bit
1. 创建分区表：
create table tb(id bigint,stat date,no bigint,pdate date,info varchar2(50)) partition by range(pdate) INTERVAL ('1 MONTH'::INTERVAL)
(
PARTITION tb_p0 VALUES LESS THAN ('2019-01-01'),
PARTITION tb_p1 VALUES LESS THAN ('2019-02-01'),
PARTITION tb_p2 VALUES LESS THAN ('2019-03-01'),
PARTITION tb_p3 VALUES LESS THAN ('2019-04-01'),
PARTITION tb_p4 VALUES LESS THAN ('2019-05-01')
);
2. 创建随机日期、数字函数：
# 随机数函数
CREATE OR REPLACE FUNCTION RANDOM_NUMBER(INTEGER, INTEGER) RETURNS INTEGER AS
$BODY$
DECLARE
START_INT ALIAS FOR $1;
END_INT ALIAS FOR $2;
BEGIN
RETURN trunc(random() * (END_INT - START_INT + 1) + START_INT);
END;
$BODY$
LANGUAGE plpgsql;
# 随机日期函数
CREATE OR REPLACE FUNCTION RANDOM_DAYS(START_DATE DATE, END_DATE DATE) RETURNS DATE AS
$BODY$
DECLARE
INTERVAL_DAYS INTEGER;
RANDOM_DAYS INTEGER;
RANDOM_DATE DATE;
BEGIN
INTERVAL_DAYS := END_DATE - START_DATE;
RANDOM_DAYS := RANDOM_NUMBER(0, INTERVAL_DAYS);
RANDOM_DATE := START_DATE + RANDOM_DAYS;
RETURN RANDOM_DATE;
END;
$BODY$
LANGUAGE plpgsql;
3. insert插入1亿条数据：
插入1亿条数据用时34分41秒，表大小为9.6GB
test=# insert into tb select generate_series(1,100000000),clock_timestamp(),RANDOM_NUMBER(1,100000000),RANDOM_DAYS('2019-01-01'::date,'2019-05-31'::date),md5(random()::text) from dual;
INSERT 0 100000000
Time: 2081789.125 ms (34:41.789)
test=# select sys_size_pretty(sys_relation_size('tb_tb_p0'));
sys_size_pretty
-----------------
0 bytes
(1 row)
Time: 36.671 ms
test=# select sys_size_pretty(sys_relation_size('tb_tb_p1'));
sys_size_pretty
-----------------
1980 MB
(1 row)
Time: 0.648 ms
test=# select sys_size_pretty(sys_relation_size('tb_tb_p2'));
sys_size_pretty
-----------------
1788 MB
(1 row)
Time: 0.591 ms
test=# select sys_size_pretty(sys_relation_size('tb_tb_p3'));
sys_size_pretty
-----------------
1981 MB
(1 row)
Time: 0.551 ms
test=# select sys_size_pretty(sys_relation_size('tb_tb_p4'));
sys_size_pretty
-----------------
1916 MB
(1 row)
Time: 0.674 ms
test=# select sys_size_pretty(sys_relation_size('tb_p5'));
sys_size_pretty
-----------------
1980 MB
(1 row)
Time: 0.668 ms
4. 添加主键：
添加主键用时3分钟14秒
test=# alter table tb add constraint tb_pk primary key(id,no);
ALTER TABLE
Time: 194581.732 ms (03:14.582)
test=# \d+ tb
Partitioned table "public.tb"
Column | Type | Collation | Nullable | Default | Storage | Stats target | Description
--------+----------------------------+-----------+----------+---------+----------+--------------+-------------
id | bigint | | not null | | plain | |
stat | date | | | | plain | |
no | bigint | | not null | | plain | |
pdate | date | | | | plain | |
info | character varying(50 char) | | | | extended | |
Partition key: RANGE (pdate)
Range interval: INTERVAL ('0-1'::pg_catalog.interval)
Indexes:
"tb_pk" PRIMARY KEY, btree (id, no) INCLUDE (tableoid) GLOBAL
Partitions: tb_p5 FOR VALUES FROM ('2019-05-01 00:00:00') TO ('2019-06-01 00:00:00'),
tb_tb_p0 FOR VALUES FROM (MINVALUE) TO ('2019-01-01 00:00:00'),
tb_tb_p1 FOR VALUES FROM ('2019-01-01 00:00:00') TO ('2019-02-01 00:00:00'),
tb_tb_p2 FOR VALUES FROM ('2019-02-01 00:00:00') TO ('2019-03-01 00:00:00'),
tb_tb_p3 FOR VALUES FROM ('2019-03-01 00:00:00') TO ('2019-04-01 00:00:00'),
tb_tb_p4 FOR VALUES FROM ('2019-04-01 00:00:00') TO ('2019-05-01 00:00:00')
5. 创建组合索引：
创建组合索引用时4分钟08秒
# 创建local本地索引
test=# create index on tb(id,stat,no,info);
CREATE INDEX
Time: 248144.207 ms (04:08.144)
test=# \d+ tb
Partitioned table "public.tb"
Column | Type | Collation | Nullable | Default | Storage | Stats target | Description
--------+----------------------------+-----------+----------+---------+----------+--------------+-------------
id | bigint | | not null | | plain | |
stat | date | | | | plain | |
no | bigint | | not null | | plain | |
pdate | date | | | | plain | |
info | character varying(50 char) | | | | extended | |
Partition key: RANGE (pdate)
Range interval: INTERVAL ('0-1'::pg_catalog.interval)
Indexes:
"tb_pk" PRIMARY KEY, btree (id, no) INCLUDE (tableoid) GLOBAL
"tb_id_stat_no_info_idx" btree (id, stat, no, info)
Partitions: tb_p5 FOR VALUES FROM ('2019-05-01 00:00:00') TO ('2019-06-01 00:00:00'),
tb_tb_p0 FOR VALUES FROM (MINVALUE) TO ('2019-01-01 00:00:00'),
tb_tb_p1 FOR VALUES FROM ('2019-01-01 00:00:00') TO ('2019-02-01 00:00:00'),
tb_tb_p2 FOR VALUES FROM ('2019-02-01 00:00:00') TO ('2019-03-01 00:00:00'),
tb_tb_p3 FOR VALUES FROM ('2019-03-01 00:00:00') TO ('2019-04-01 00:00:00'),
tb_tb_p4 FOR VALUES FROM ('2019-04-01 00:00:00') TO ('2019-05-01 00:00:00')
test=# \d+ tb_tb_p2
Table "public.tb_tb_p2"
Column | Type | Collation | Nullable | Default | Storage | Stats target | Description
--------+----------------------------+-----------+----------+---------+----------+--------------+-------------
id | bigint | | not null | | plain | |
stat | date | | | | plain | |
no | bigint | | not null | | plain | |
pdate | date | | | | plain | |
info | character varying(50 char) | | | | extended | |
Partition of: tb FOR VALUES FROM ('2019-02-01 00:00:00') TO ('2019-03-01 00:00:00')
Partition constraint: (((pdate IS NULL) OR ((pdate)::timestamp without time zone >= '2019-02-01 00:00:00'::date)) AND ((pdate IS NOT NULL) AND ((pdate)::timestamp without time zone < '2019-03-01 00:00:00'::date)))
Indexes:
"tb_tb_p2_id_stat_no_info_idx" btree (id, stat, no, info)
Access method: heap
# 使用global创建全局索引
# 查询表信息，发现使用global声明创建的索引也是local本地索引
test=# create index on tb(id,stat) global;
CREATE INDEX
Time: 151558.143 ms (02:31.558)
test=# \d+ tb
Partitioned table "public.tb"
Column | Type | Collation | Nullable | Default | Storage | Stats target | Description
--------+----------------------------+-----------+----------+---------+----------+--------------+-------------
id | bigint | | not null | | plain | |
stat | date | | | | plain | |
no | bigint | | not null | | plain | |
pdate | date | | | | plain | |
info | character varying(50 char) | | | | extended | |
Partition key: RANGE (pdate)
Range interval: INTERVAL ('0-1'::pg_catalog.interval)
Indexes:
"tb_pk" PRIMARY KEY, btree (id, no) INCLUDE (tableoid) GLOBAL
"tb_id_stat_idx" btree (id, stat)
Partitions: tb_p5 FOR VALUES FROM ('2019-05-01 00:00:00') TO ('2019-06-01 00:00:00'),
tb_tb_p0 FOR VALUES FROM (MINVALUE) TO ('2019-01-01 00:00:00'),
tb_tb_p1 FOR VALUES FROM ('2019-01-01 00:00:00') TO ('2019-02-01 00:00:00'),
tb_tb_p2 FOR VALUES FROM ('2019-02-01 00:00:00') TO ('2019-03-01 00:00:00'),
tb_tb_p3 FOR VALUES FROM ('2019-03-01 00:00:00') TO ('2019-04-01 00:00:00'),
tb_tb_p4 FOR VALUES FROM ('2019-04-01 00:00:00') TO ('2019-05-01 00:00:00')
test=# \d+ tb_tb_p4
Table "public.tb_tb_p4"
Column | Type | Collation | Nullable | Default | Storage | Stats target | Description
--------+----------------------------+-----------+----------+---------+----------+--------------+-------------
id | bigint | | not null | | plain | |
stat | date | | | | plain | |
no | bigint | | not null | | plain | |
pdate | date | | | | plain | |
info | character varying(50 char) | | | | extended | |
Partition of: tb FOR VALUES FROM ('2019-04-01 00:00:00') TO ('2019-05-01 00:00:00')
Partition constraint: (((pdate IS NULL) OR ((pdate)::timestamp without time zone >= '2019-04-01 00:00:00'::date)) AND ((pdate IS NOT NULL) AND ((pdate)::timestamp without time zone < '2019-05-01 00:00:00'::date)))
Indexes:
"tb_tb_p4_id_stat_idx" btree (id, stat)
Access method: heap
# 使用unique加分区键pdate创建全局索引（global声明）
# 查询表信息，创建的索引依然是local本地索引
test=# create unique index ON tb(id,pdate) global;
CREATE INDEX
Time: 87232.640 ms (01:27.233)
test=# \d+ tb
Partitioned table "public.tb"
Column | Type | Collation | Nullable | Default | Storage | Stats target | Description
--------+----------------------------+-----------+----------+---------+----------+--------------+-------------
id | bigint | | | | plain | |
stat | date | | | | plain | |
no | bigint | | | | plain | |
pdate | date | | | | plain | |
info | character varying(50 char) | | | | extended | |
Partition key: RANGE (pdate)
Range interval: INTERVAL ('0-1'::pg_catalog.interval)
Indexes:
"tb_id_pdate_idx" UNIQUE, btree (id, pdate)
Partitions: tb_p5 FOR VALUES FROM ('2019-05-01 00:00:00') TO ('2019-06-01 00:00:00'),
tb_tb_p0 FOR VALUES FROM (MINVALUE) TO ('2019-01-01 00:00:00'),
tb_tb_p1 FOR VALUES FROM ('2019-01-01 00:00:00') TO ('2019-02-01 00:00:00'),
tb_tb_p2 FOR VALUES FROM ('2019-02-01 00:00:00') TO ('2019-03-01 00:00:00'),
tb_tb_p3 FOR VALUES FROM ('2019-03-01 00:00:00') TO ('2019-04-01 00:00:00'),
tb_tb_p4 FOR VALUES FROM ('2019-04-01 00:00:00') TO ('2019-05-01 00:00:00')
# 使用unique加非分区键创建全局索引（global声明）
# 查询表信息，全局索引创建成功
test=# create unique index ON tb(id,info) global;
CREATE INDEX
test=# \d+ tb
Partitioned table "public.tb"
Column | Type | Collation | Nullable | Default | Storage | Stats target | Description
--------+----------------------------+-----------+----------+---------+----------+--------------+-------------
id | bigint | | | | plain | |
stat | date | | | | plain | |
no | bigint | | | | plain | |
pdate | date | | | | plain | |
info | character varying(50 char) | | | | extended | |
Partition key: RANGE (pdate)
Range interval: INTERVAL ('0-1'::pg_catalog.interval)
Indexes:
"tb_id_info_idx" UNIQUE, btree (id, info) INCLUDE (tableoid) GLOBAL
Partitions: tb_p5 FOR VALUES FROM ('2019-05-01 00:00:00') TO ('2019-06-01 00:00:00'),
tb_tb_p0 FOR VALUES FROM (MINVALUE) TO ('2019-01-01 00:00:00'),
tb_tb_p1 FOR VALUES FROM ('2019-01-01 00:00:00') TO ('2019-02-01 00:00:00'),
tb_tb_p2 FOR VALUES FROM ('2019-02-01 00:00:00') TO ('2019-03-01 00:00:00'),
tb_tb_p3 FOR VALUES FROM ('2019-03-01 00:00:00') TO ('2019-04-01 00:00:00'),
tb_tb_p4 FOR VALUES FROM ('2019-04-01 00:00:00') TO ('2019-05-01 00:00:00')
总结：
1 添加主键的同时会创建主键列（字段）唯一索引（但是有唯一索引的列不一定是主键）。
2 主键字段不允许空值，添加主键过程中会自动添加not null非空约束，保证主键列值的唯一性。
3 分区表添加主键同时创建的索引（索引有GLOBAL）是全局索引。
4 在分区表创建索引（默认为local本地索引，全局索引需要global声明），会自动在每个分区上创建一个本地索引。
5 分区表会在主键列创建一个全局（global）索引，默认为添加主键列的同时创建全局索引。
6 分区表惟一约束必须包含分区键。
7 分区表成功创建全局索引必须满足条件：索引类型是唯一索引（unique）并且不包含分区键 。
三、设置数据库参数创建主键与索引
1. 添加主键：
先设置字段非空not null约束，然后再添加主键。效率有提升。
test=# alter table tb alter column id set not null;
ALTER TABLE
Time: 37.610 ms
test=# alter table tb alter column no set not null;
ALTER TABLE
Time: 3.629 ms
test=# alter table tb add constraint tb_pk primary key(id,no);
ALTER TABLE
Time: 161859.979 ms (02:41.860)
test=# \d+ tb
Partitioned table "public.tb"
Column | Type | Collation | Nullable | Default | Storage | Stats target | Description
--------+----------------------------+-----------+----------+---------+----------+--------------+-------------
id | bigint | | not null | | plain | |
stat | date | | | | plain | |
no | bigint | | not null | | plain | |
pdate | date | | | | plain | |
info | character varying(50 char) | | | | extended | |
Partition key: RANGE (pdate)
Range interval: INTERVAL ('0-1'::pg_catalog.interval)
Indexes:
"tb_pk" PRIMARY KEY, btree (id, no) INCLUDE (tableoid) GLOBAL
Partitions: tb_p5 FOR VALUES FROM ('2019-05-01 00:00:00') TO ('2019-06-01 00:00:00'),
tb_tb_p0 FOR VALUES FROM (MINVALUE) TO ('2019-01-01 00:00:00'),
tb_tb_p1 FOR VALUES FROM ('2019-01-01 00:00:00') TO ('2019-02-01 00:00:00'),
tb_tb_p2 FOR VALUES FROM ('2019-02-01 00:00:00') TO ('2019-03-01 00:00:00'),
tb_tb_p3 FOR VALUES FROM ('2019-03-01 00:00:00') TO ('2019-04-01 00:00:00'),
tb_tb_p4 FOR VALUES FROM ('2019-04-01 00:00:00') TO ('2019-05-01 00:00:00')
2. 设置work_mem、maintenance_work_mem、temp_buffers参数然后添加主键：
设置数据库参数、添加非空not null约束后添加主键用时2分50秒
test=# set work_mem ='100MB';
SET
test=# set maintenance_work_mem ='500MB';
SET
test=# set temp_buffers ='100MB';
SET
# 然后设置非空约束not null添加主键
test=# alter table tb alter column id set not null;
ALTER TABLE
Time: 20.714 ms
test=# alter table tb alter column no set not null;
ALTER TABLE
Time: 2.202 ms
test=# alter table tb add constraint tb_pk primary key(id,no);
ALTER TABLE
Time: 170640.065 ms (02:50.640)
test=# \d+ tb
Partitioned table "public.tb"
Column | Type | Collation | Nullable | Default | Storage | Stats target | Description
--------+----------------------------+-----------+----------+---------+----------+--------------+-------------
id | bigint | | not null | | plain | |
stat | date | | | | plain | |
no | bigint | | not null | | plain | |
pdate | date | | | | plain | |
info | character varying(50 char) | | | | extended | |
Partition key: RANGE (pdate)
Range interval: INTERVAL ('0-1'::pg_catalog.interval)
Indexes:
"tb_pk" PRIMARY KEY, btree (id, no) INCLUDE (tableoid) GLOBAL
Partitions: tb_p5 FOR VALUES FROM ('2019-05-01 00:00:00') TO ('2019-06-01 00:00:00'),
tb_tb_p0 FOR VALUES FROM (MINVALUE) TO ('2019-01-01 00:00:00'),
tb_tb_p1 FOR VALUES FROM ('2019-01-01 00:00:00') TO ('2019-02-01 00:00:00'),
tb_tb_p2 FOR VALUES FROM ('2019-02-01 00:00:00') TO ('2019-03-01 00:00:00'),
tb_tb_p3 FOR VALUES FROM ('2019-03-01 00:00:00') TO ('2019-04-01 00:00:00'),
tb_tb_p4 FOR VALUES FROM ('2019-04-01 00:00:00') TO ('2019-05-01 00:00:00')
通过以上设置数据库参数，分区表添加非空not null约束可以看出，大数据量的分区表添加主键性能是有明显提升。
3. 创建索引：
# 创建本地local索引
test=# create index on tb(id,stat,no,info);
CREATE INDEX
Time: 250377.917 ms (04:10.378)
test=# \d+ tb
Partitioned table "public.tb"
Column | Type | Collation | Nullable | Default | Storage | Stats target | Description
--------+----------------------------+-----------+----------+---------+----------+--------------+-------------
id | bigint | | not null | | plain | |
stat | date | | | | plain | |
no | bigint | | not null | | plain | |
pdate | date | | | | plain | |
info | character varying(50 char) | | | | extended | |
Partition key: RANGE (pdate)
Range interval: INTERVAL ('0-1'::pg_catalog.interval)
Indexes:
"tb_pk" PRIMARY KEY, btree (id, no) INCLUDE (tableoid) GLOBAL
"tb_id_stat_no_info_idx" btree (id, stat, no, info)
Partitions: tb_p5 FOR VALUES FROM ('2019-05-01 00:00:00') TO ('2019-06-01 00:00:00'),
tb_tb_p0 FOR VALUES FROM (MINVALUE) TO ('2019-01-01 00:00:00'),
tb_tb_p1 FOR VALUES FROM ('2019-01-01 00:00:00') TO ('2019-02-01 00:00:00'),
tb_tb_p2 FOR VALUES FROM ('2019-02-01 00:00:00') TO ('2019-03-01 00:00:00'),
tb_tb_p3 FOR VALUES FROM ('2019-03-01 00:00:00') TO ('2019-04-01 00:00:00'),
tb_tb_p4 FOR VALUES FROM ('2019-04-01 00:00:00') TO ('2019-05-01 00:00:00')
# 使用global创建全局索引
test=# create index on tb(id,stat) global;
CREATE INDEX
Time: 160475.453 ms (02:40.475)
test=# \d+ tb
Partitioned table "public.tb"
Column | Type | Collation | Nullable | Default | Storage | Stats target | Description
--------+----------------------------+-----------+----------+---------+----------+--------------+-------------
id | bigint | | not null | | plain | |
stat | date | | | | plain | |
no | bigint | | not null | | plain | |
pdate | date | | | | plain | |
info | character varying(50 char) | | | | extended | |
Partition key: RANGE (pdate)
Range interval: INTERVAL ('0-1'::pg_catalog.interval)
Indexes:
"tb_pk" PRIMARY KEY, btree (id, no) INCLUDE (tableoid) GLOBAL
"tb_id_stat_idx" btree (id, stat)
Partitions: tb_p5 FOR VALUES FROM ('2019-05-01 00:00:00') TO ('2019-06-01 00:00:00'),
tb_tb_p0 FOR VALUES FROM (MINVALUE) TO ('2019-01-01 00:00:00'),
tb_tb_p1 FOR VALUES FROM ('2019-01-01 00:00:00') TO ('2019-02-01 00:00:00'),
tb_tb_p2 FOR VALUES FROM ('2019-02-01 00:00:00') TO ('2019-03-01 00:00:00'),
tb_tb_p3 FOR VALUES FROM ('2019-03-01 00:00:00') TO ('2019-04-01 00:00:00'),
tb_tb_p4 FOR VALUES FROM ('2019-04-01 00:00:00') TO ('2019-05-01 00:00:00')
从以上结果可以看出，设置数据库参数、分区表非空not null约束，添加主键的性能有提升。但是创建索引的性能无明显提升。
四、其他方式
1. 使用on only创建索引：
优点：创建索引时不会阻止对分区表的并发插入、更新或者删除操作。
缺点：分区表父表不支持CONCURRENTLY、parallel_workers选项，子分区支持CONCURRENTLY、parallel_workers选项。
1 使用on only创建失效invalid索引，使所有的子分区不会自动应用该索引：
test=# create index tb_idx on only tb(id,stat,no,info);
CREATE INDEX
Time: 21.135 ms
test=# \d+ tb
Partitioned table "public.tb"
Column | Type | Collation | Nullable | Default | Storage | Stats target | Description
--------+----------------------------+-----------+----------+---------+----------+--------------+-------------
id | bigint | | not null | | plain | |
stat | date | | | | plain | |
no | bigint | | not null | | plain | |
pdate | date | | | | plain | |
info | character varying(50 char) | | | | extended | |
Partition key: RANGE (pdate)
Range interval: INTERVAL ('0-1'::pg_catalog.interval)
Indexes:
"tb_pk" PRIMARY KEY, btree (id, no) INCLUDE (tableoid) GLOBAL
"tb_idx" btree (id, stat, no, info) INVALID
Partitions: tb_p5 FOR VALUES FROM ('2019-05-01 00:00:00') TO ('2019-06-01 00:00:00'),
tb_tb_p0 FOR VALUES FROM (MINVALUE) TO ('2019-01-01 00:00:00'),
tb_tb_p1 FOR VALUES FROM ('2019-01-01 00:00:00') TO ('2019-02-01 00:00:00'),
tb_tb_p2 FOR VALUES FROM ('2019-02-01 00:00:00') TO ('2019-03-01 00:00:00'),
tb_tb_p3 FOR VALUES FROM ('2019-03-01 00:00:00') TO ('2019-04-01 00:00:00'),
tb_tb_p4 FOR VALUES FROM ('2019-04-01 00:00:00') TO ('2019-05-01 00:00:00')
2 然后手动创建分区表子分区索引：
# 手动给分区表子分区创建索引
test=# create index tb_tb_p0_idx on tb_tb_p0(id,stat,no,info);
CREATE INDEX
Time: 5.469 ms
test=# create index tb_tb_p1_idx on tb_tb_p1(id,stat,no,info);
CREATE INDEX
Time: 50423.600 ms (00:50.424)
test=# create index tb_tb_p2_idx on tb_tb_p2(id,stat,no,info);
CREATE INDEX
Time: 46672.063 ms (00:46.672)
test=# create index tb_tb_p3_idx on tb_tb_p3(id,stat,no,info);
CREATE INDEX
Time: 50621.904 ms (00:50.622)
test=# create index tb_tb_p4_idx on tb_tb_p4(id,stat,no,info);
CREATE INDEX
Time: 49248.159 ms (00:49.248)
test=# create index tb_p5_idx on tb_p5(id,stat,no,info);
CREATE INDEX
Time: 52901.838 ms (00:52.902)
# 然后使用ALTER INDEX .. ATTACH PARTITION到父索引
test=# alter index tb_idx attach partition tb_tb_p0_idx;
ALTER INDEX
Time: 5.326 ms
test=# alter index tb_idx attach partition tb_tb_p1_idx;
ALTER INDEX
Time: 2.688 ms
test=# alter index tb_idx attach partition tb_tb_p2_idx;
ALTER INDEX
Time: 2.908 ms
test=# alter index tb_idx attach partition tb_tb_p3_idx;
ALTER INDEX
Time: 3.608 ms
test=# alter index tb_idx attach partition tb_tb_p4_idx;
ALTER INDEX
Time: 2.332 ms
test=# alter index tb_idx attach partition tb_p5_idx;
ALTER INDEX
Time: 37.518 ms
# 所有的子分区索引都附加到父索引后，父索引将自动标记为有效
test=# \d+ tb
Partitioned table "public.tb"
Column | Type | Collation | Nullable | Default | Storage | Stats target | Description
--------+----------------------------+-----------+----------+---------+----------+--------------+-------------
id | bigint | | not null | | plain | |
stat | date | | | | plain | |
no | bigint | | not null | | plain | |
pdate | date | | | | plain | |
info | character varying(50 char) | | | | extended | |
Partition key: RANGE (pdate)
Range interval: INTERVAL ('0-1'::pg_catalog.interval)
Indexes:
"tb_pk" PRIMARY KEY, btree (id, no) INCLUDE (tableoid) GLOBAL
"tb_idx" btree (id, stat, no, info)
Partitions: tb_p5 FOR VALUES FROM ('2019-05-01 00:00:00') TO ('2019-06-01 00:00:00'),
tb_tb_p0 FOR VALUES FROM (MINVALUE) TO ('2019-01-01 00:00:00'),
tb_tb_p1 FOR VALUES FROM ('2019-01-01 00:00:00') TO ('2019-02-01 00:00:00'),
tb_tb_p2 FOR VALUES FROM ('2019-02-01 00:00:00') TO ('2019-03-01 00:00:00'),
tb_tb_p3 FOR VALUES FROM ('2019-03-01 00:00:00') TO ('2019-04-01 00:00:00'),
tb_tb_p4 FOR VALUES FROM ('2019-04-01 00:00:00') TO ('2019-05-01 00:00:00')
test=# \d+ tb_tb_p3
Table "public.tb_tb_p3"
Column | Type | Collation | Nullable | Default | Storage | Stats target | Description
--------+----------------------------+-----------+----------+---------+----------+--------------+-------------
id | bigint | | not null | | plain | |
stat | date | | | | plain | |
no | bigint | | not null | | plain | |
pdate | date | | | | plain | |
info | character varying(50 char) | | | | extended | |
Partition of: tb FOR VALUES FROM ('2019-03-01 00:00:00') TO ('2019-04-01 00:00:00')
Partition constraint: (((pdate IS NULL) OR ((pdate)::timestamp without time zone >= '2019-03-01 00:00:00'::date)) AND ((pdate IS NOT NULL) AND ((pdate)::timestamp without time zone < '2019-04-01 00:00:00'::date)))
Indexes:
"tb_tb_p3_idx" btree (id, stat, no, info)
Access method: heap
总结：
分区表父表不支持CONCURRENTLY、parallel_workers选项，但是子分区支持CONCURRENTLY、parallel_workers选项。
可以使用on only方式在父表创建失效的索引，设置子分区表CONCURRENTLY、parallel_workers选项，手动创建子分区索引，然后使用ALTER INDEX .. ATTACH PARTITION到父索引，父索引会自动标记为有效。
同时分区表也不支持以下添加主键语法：
create unique index tb_idx on tb(id);
alter table tb ADD CONSTRAINT tb_pk PRIMARY KEY using index tb_idx;
KINGBASE研究院
