Help Center/ GaussDB/ M-Compatibility Developer Guide(Distributed)/ SQL Reference/ SQL Syntax/ SQL Statements/ R/ REINDEX
Updated on 2025-10-23 GMT+08:00
View PDF
Share

REINDEX

Description

Rebuilds an index using the data stored in the index's table, replacing the old copy of the index.

There are several scenarios in which REINDEX can be used:

  • An index has become corrupted, and no longer contains valid data.
  • An index has become "bloated", that is, it contains many empty or nearly-empty pages.
  • You have altered a storage parameter (such as a fill factor) for an index, and wish that the change takes full effect.

Precautions

  • REINDEX DATABASE and REINDEX SYSTEM cannot be performed in transaction blocks.
  • If the index contains the lpi_parallel_method option and the value is PARTITION, and the parallel_workers value of the index's table is greater than 0, the index cannot be rebuilt in parallel. If the index does not contain the lpi_parallel_method option or the value of the option is set to AUTO, page-level parallel index is rebuilt by default. For details, see LPI_PARALLEL_METHOD.

Syntax

  • Rebuild an ordinary index. 
    REINDEX { INDEX |  TABLE | DATABASE} [CONCURRENTLY] name [ FORCE ];
REINDEX SYSTEM name [FORCE];
  • Rebuild an index partition. 
    REINDEX { INDEX | TABLE }[CONCURRENTLY] name PARTITION partition_name [ FORCE ];

Parameters

  • INDEX

    Rebuilds the specified index.

  • TABLE

    Rebuilds all indexes of a specified table. The TOAST table (if any) of the table is reindexed as well. If an index on the table has been invalidated by running alter unusable, the index cannot be rebuilt. Indexes in the TOAST table cannot be rebuilt when specifying the CONCURRENTLY option.

  • DATABASE

    Rebuilds all indexes within the current database. Indexes in the TOAST table within the current database cannot be rebuilt when specifying the CONCURRENTLY option.

  • SYSTEM

    Rebuilds all indexes on system catalogs within the current database. Indexes on user tables are not processed.

  • CONCURRENTLY

    Rebuilds an index (with ShareUpdateExclusiveLock) in non-blocking DML mode. When an index is rebuilt, other statements cannot access the table on which the index depends. If this keyword is specified, DML is not blocked during the recreation. Indexes in system catalogs cannot be rebuilt online. REINDEX INTERNAL TABLE CONCURRENTLY, REINDEX SYSTEM CONCURRENTLY, and REINDEX INVALID INDEX CONCURRENTLY are not supported. When REINDEX DATABASE CONCURRENTLY is executed, all indexes on user tables in the current database are rebuilt online (indexes on system catalogs are not processed). REINDEX CONCURRENTLY cannot be executed within a transaction. Online index rebuilding supports only B-tree and UB-tree indexes, ordinary indexes, global indexes, and local indexes. Online index columns cannot be added, deleted, or modified. PCR UB-tree indexes, level-2 partitions, and GSIs are not supported. Online concurrent index rebuilding supports only ordinary indexes, global indexes, and local indexes of Astore and Ustore. Other online index rebuilding specifications are inherited from the current version. If online index rebuilding fails, the system automatically clears new indexes to prevent resource occupation in scenarios such as manual cancellation, duplicate unique index key values, insufficient resources, thread startup failure, and lock timeout. If the system cannot automatically clear invalid new indexes (for example, the database breaks down, FATAL, or PANIC), you need to manually clear invalid new indexes (using the DROP INDEX statement) and temporary tables (using the DROP TABLE statement) as soon as possible to prevent more resources from being occupied. Generally, the extension of an invalid index name is _ccnew. The execution of REINDEX INDEX CONCURRENTLY adds a four-level session lock to the table and its first several phases are similar to those of CREATE INDEX CONCURRENTLY. Therefore, the execution may be suspended or deadlocked, which is similar to that of CREATE INDEX CONCURRENTLY. For example, if two sessions perform the REINDEX CONCURRENTLY operation on the same index or table at the same time, a deadlock occurs.

    This keyword is specified when an index is rebuilt. For Astore, you need to complete two full table scans for building. During the first scan, a new index is created without blocking read and write operations. During the second scan, the changes in the first scan are merged and updated. For Ustore, you need to complete a full table scan. During the scan, data generated by concurrent DML operations is inserted into the temporary table named index_oid_cctmp. After the scan is complete, you merge the temporary table to the new index suffixed with _ccnew{n}, delete the temporary table, exchange the old and new indexes, mark the old index as dead, enable the new index, and rebuild the index.

  • name

    Specifies the name of the index, table, or database whose index needs to be rebuilt. Tables and indexes can be schema-qualified.

    REINDEX DATABASE and SYSTEM can create indexes for only the current database. Therefore, name must be the same as the current database name.

  • FORCE

    Specifies an invalid option, which will be ignored.

  • partition_name

    Specifies the name of the partition or index partition to be rebuilt.

    Value range:

    • If REINDEX INDEX is used, specify the name of an index partition.
    • If REINDEX TABLE is used, specify the name of a partition.

REINDEX DATABASE and REINDEX SYSTEM cannot be performed in transaction blocks.

Examples

-- Create a schema.
m_db=# CREATE SCHEMA tpcds;

-- Create the tpcds. customer table.
m_db=# CREATE TABLE tpcds.customer
(
c_customer_sk         INTEGER        NOT NULL,
c_customer_id         CHAR(16)       NOT NULL
);

-- Insert multiple records into the table.
m_db=# INSERT INTO tpcds.customer VALUES (1, 'AAAAAAAABAAAAAAA'),(5, 'AAAAAAAACAAAAAAA'),(10, 'AAAAAAAADAAAAAAA');

-- Create a row-store table tpcds.customer_t1 and create an index on the c_customer_sk column in the table.
m_db=# CREATE TABLE tpcds.customer_t1
(
    c_customer_sk             integer               not null,
    c_customer_id             char(16)              not null,
    c_current_cdemo_sk        integer                       ,
    c_current_hdemo_sk        integer                       ,
    c_current_addr_sk         integer                       ,
    c_first_shipto_date_sk    integer                       ,
    c_first_sales_date_sk     integer                       ,
    c_salutation              char(10)                      ,
    c_first_name              char(20)                      ,
    c_last_name               char(30)                      ,
    c_preferred_cust_flag     char(1)                       ,
    c_birth_day               integer                       ,
    c_birth_month             integer                       ,
    c_birth_year              integer                       ,
    c_birth_country           varchar(20)                   ,
    c_login                   char(13)                      ,
    c_email_address           char(50)                      ,
    c_last_review_date        char(10)
)
WITH (orientation = row);

m_db=# CREATE INDEX tpcds_customer_index1 ON tpcds.customer_t1 (c_customer_sk);

m_db=# INSERT INTO tpcds.customer_t1 SELECT * FROM tpcds.customer WHERE c_customer_sk < 10;

-- Rebuild a single index.
m_db=# REINDEX INDEX tpcds.tpcds_customer_index1;

-- Rebuild all indexes in the tpcds.customer_t1 table:
m_db=# REINDEX TABLE tpcds.customer_t1;

Delete the tpcds.customer_t1 table.
m_db=# DROP TABLE tpcds.customer_t1;

-- Drop the table.
m_db=# DROP TABLE tpcds.customer;

-- Delete the schema.
m_db=# DROP SCHEMA tpcds;

Suggestions

  • INTERNAL TABLE

    This scenario is used for fault recovery. You are advised not to perform concurrent operations.

  • DATABASE

    You are not allowed to reindex databases in transactions.

  • SYSTEM

    You are not allowed to reindex system catalogs in transactions.

Parent topic: R