ATTACH Statement

Documentation for Attach

Attaches a table or a dictionary, for example, when moving a database to another server.

Syntax

ATTACH TABLE|DICTIONARY|DATABASE [IF NOT EXISTS] [db.]name [ON CLUSTER cluster] ...

The query does not create data on disk, but assumes that data is already in the appropriate places, and just adds information about the specified table, dictionary or database to the server. After executing the ATTACH query, the server will know about the existence of the table, dictionary or database.

If a table was previously detached (DETACH query), meaning that its structure is known, you can use shorthand without defining the structure.

Attach Existing Table

Syntax

ATTACH TABLE [IF NOT EXISTS] [db.]name [ON CLUSTER cluster]

This query is used when starting the server. The server stores table metadata as files with ATTACH queries, which it simply runs at launch (with the exception of some system tables, which are explicitly created on the server).

If the table was detached permanently, it won’t be reattached at the server start, so you need to use ATTACH query explicitly.

Create New Table And Attach Data

With Specified Path to Table Data

The query creates a new table with provided structure and attaches table data from the provided directory in user_files.

Syntax

ATTACH TABLE name FROM 'path/to/data/' (col1 Type1, ...)

Example

Query:

DROP TABLE IF EXISTS test;
INSERT INTO TABLE FUNCTION file('01188_attach/test/data.TSV', 'TSV', 's String, n UInt8') VALUES ('test', 42);
ATTACH TABLE test FROM '01188_attach/test' (s String, n UInt8) ENGINE = File(TSV);
SELECT * FROM test;

Result:

┌─s────┬──n─┐
│ test │ 42 │
└──────┴────┘

With Specified Table UUID

This query creates a new table with provided structure and attaches data from the table with the specified UUID. It is supported by the Atomic database engine.

Syntax

ATTACH TABLE name UUID '<uuid>' (col1 Type1, ...)

Attach MergeTree table as ReplicatedMergeTree

Allows to attach non-replicated MergeTree table as ReplicatedMergeTree. ReplicatedMergeTree table will be created with values of default_replica_path and default_replica_name settings. It is also possible to attach a replicated table as a regular MergeTree.

Note that table’s data in ZooKeeper is not affected in this query. This means you have to add metadata in ZooKeeper using SYSTEM RESTORE REPLICA or clear it with SYSTEM DROP REPLICA ... FROM ZKPATH ... after attach.

If you are trying to add a replica to an existing ReplicatedMergeTree table, keep in mind that all the local data in converted MergeTree table will be detached.

Syntax

ATTACH TABLE [db.]name AS [NOT] REPLICATED

Convert table to replicated

DETACH TABLE test;
ATTACH TABLE test AS REPLICATED;
SYSTEM RESTORE REPLICA test;

Convert table to not replicated

Get ZooKeeper path and replica name for table:

SELECT replica_name, zookeeper_path FROM system.replicas WHERE table='test';

Result:

┌─replica_name─┬─zookeeper_path─────────────────────────────────────────────┐
│ r1           │ /clickhouse/tables/401e6a1f-9bf2-41a3-a900-abb7e94dff98/s1 │
└──────────────┴────────────────────────────────────────────────────────────┘

Attach table as not replicated and delete replica’s data from ZooKeeper:

DETACH TABLE test;
ATTACH TABLE test AS NOT REPLICATED;
SYSTEM DROP REPLICA 'r1' FROM ZKPATH '/clickhouse/tables/401e6a1f-9bf2-41a3-a900-abb7e94dff98/s1';

Attach Existing Dictionary

Attaches a previously detached dictionary.

Syntax

ATTACH DICTIONARY [IF NOT EXISTS] [db.]name [ON CLUSTER cluster]

Attach Existing Database

Attaches a previously detached database.

Syntax

ATTACH DATABASE [IF NOT EXISTS] name [ENGINE=<database engine>] [ON CLUSTER cluster]