PG建立or修改索引

pgsql中的索引不能重名，重名的建立失敗。

 

建立二級索引的命令：create index CONCURRENTLY idx_abc on tb1(a,b);

 

注意：reindex 重建索引的過程是阻塞的，通常大表不建議使用這個命令，能夠重建一個索引，而後刪除老的索引。

 

下面看一個《PostgreSQL實戰》書上page202的例子：

因爲PG的MVCC機制，當運行大量的更新操做後，會有索引膨脹的現象。這時候 能夠經過 create index concurrently 不阻塞查詢和更新的狀況下，在線從新重建索引，建立好索引以後，再刪除原先的索引，減小索引的尺寸，提升查詢速度。對於主鍵也能夠使用這種方式進行，例如：

db1=# \d testdata

                 Table "public.testdata"

  Column  |     Type     | Collation | Nullable | Default

----------+--------------+-----------+----------+---------

 id       | integer      |           | not null |

 course   | integer      |           |          |

 grade    | numeric(4,2) |           |          |

 testtime | date         |           |          |

Indexes:

    "testdata_pkey" PRIMARY KEY, btree (id)

 

db1=# create unique index concurrently on testdata using btree(id);

 

db1=# select

schemaname,

relname,

indexrelname,

pg_relation_size (indexrelid) as index_size,

idx_scan,

idx_tup_read,

idx_tup_fetch

from pg_stat_user_indexes where

indexrelname in (select indexname from pg_indexes where schemaname ='public' and tablename='testdata');

 schemaname | relname  |  indexrelname   | index_size | idx_scan | idx_tup_read | idx_tup_fetch

------------+----------+-----------------+------------+----------+--------------+---------------

 public     | testdata | testdata_pkey   |      16384 |        2 |      5999998 |       5999998

 public     | testdata | testdata_id_idx |      16384 |        0 |            0 |             0

(2 rows)

 

db1=# begin;

BEGIN

 

db1=# alter table testdata drop constraint testdata_pkey;

ALTER TABLE

 

db1=# alter table testdata add constraint testdata_id_idx primary key using index testdata_id_idx ;

ALTER TABLE

 

db1=# end;

COMMIT

 

db1=# \d testdata

                 Table "public.testdata"

  Column  |     Type     | Collation | Nullable | Default

----------+--------------+-----------+----------+---------

 id       | integer      |           | not null |

 course   | integer      |           |          |

 grade    | numeric(4,2) |           |          |

 testtime | date         |           |          |

Indexes:

    "testdata_id_idx" PRIMARY KEY, btree (id)

 

db1=# select

schemaname,

relname,

indexrelname,

pg_relation_size (indexrelid) as index_size,

idx_scan,

idx_tup_read,

idx_tup_fetch

from pg_stat_user_indexes where

indexrelname in (select indexname from pg_indexes where schemaname ='public' and tablename='testdata');

schemaname | relname  |  indexrelname   | index_size | idx_scan | idx_tup_read | idx_tup_fetch

------------+----------+-----------------+------------+----------+--------------+---------------

 public     | testdata | testdata_id_idx |      16384 |        0 |            0 |             0

(1 row)

 

 

這樣就完成了主鍵索引的重建，對於大規模的數據庫集羣，能夠經過 pg_repack 工具進行定時的索引重建。