ClickHouse/docs/en/query_language/alter.md

22 KiB
Raw Blame History

ALTER

The ALTER query is only supported for *MergeTree tables, as well as MergeandDistributed. The query has several variations.

Column Manipulations

Changing the table structure.

ALTER TABLE [db].name [ON CLUSTER cluster] ADD|DROP|CLEAR|COMMENT|MODIFY COLUMN ...

In the query, specify a list of one or more comma-separated actions. Each action is an operation on a column.

The following actions are supported:

Detailed description of these actions is shown below.

ADD COLUMN

ADD COLUMN [IF NOT EXISTS] name [type] [default_expr] [AFTER name_after]

Adds a new column to the table with the specified name, type, and default_expr (see the section Default expressions).

If the IF NOT EXISTS clause is included, the query won't return an error if the column already exists. If you specify AFTER name_after (the name of another column), the column is added after the specified one in the list of table columns. Otherwise, the column is added to the end of the table. Note that there is no way to add a column to the beginning of a table. For a chain of actions, name_after can be the name of a column that is added in one of the previous actions.

Adding a column just changes the table structure, without performing any actions with data. The data doesn't appear on the disk after ALTER. If the data is missing for a column when reading from the table, it is filled in with default values (by performing the default expression if there is one, or using zeros or empty strings). The column appears on the disk after merging data parts (see MergeTree).

This approach allows us to complete the ALTER query instantly, without increasing the volume of old data.

Example:

ALTER TABLE visits ADD COLUMN browser String AFTER user_id

DROP COLUMN

DROP COLUMN [IF EXISTS] name

Deletes the column with the name name. If the IF EXISTS clause is specified, the query won't return an error if the column doesn't exist.

Deletes data from the file system. Since this deletes entire files, the query is completed almost instantly.

Example:

ALTER TABLE visits DROP COLUMN browser

CLEAR COLUMN

CLEAR COLUMN [IF EXISTS] name IN PARTITION partition_name

Resets all data in a column for a specified partition. Read more about setting the partition name in the section How to specify the partition expression.

If the IF EXISTS clause is specified, the query won't return an error if the column doesn't exist.

Example:

ALTER TABLE visits CLEAR COLUMN browser IN PARTITION tuple()

COMMENT COLUMN

COMMENT COLUMN [IF EXISTS] name 'comment'

Adds a comment to the column. If the IF EXISTS clause is specified, the query won't return an error if the column doesn't exist.

Each column can have one comment. If a comment already exists for the column, a new comment overwrites the previous comment.

Comments are stored in the comment_expression column returned by the DESCRIBE TABLE query.

Example:

ALTER TABLE visits COMMENT COLUMN browser 'The table shows the browser used for accessing the site.'

MODIFY COLUMN

MODIFY COLUMN [IF EXISTS] name [type] [default_expr]

This query changes the name column's type to type and/or the default expression to default_expr. If the IF EXISTS clause is specified, the query won't return an error if the column doesn't exist.

When changing the type, values are converted as if the toType functions were applied to them. If only the default expression is changed, the query doesn't do anything complex, and is completed almost instantly.

Example:

ALTER TABLE visits MODIFY COLUMN browser Array(String)

Changing the column type is the only complex action it changes the contents of files with data. For large tables, this may take a long time.

There are several processing stages:

  • Preparing temporary (new) files with modified data.
  • Renaming old files.
  • Renaming the temporary (new) files to the old names.
  • Deleting the old files.

Only the first stage takes time. If there is a failure at this stage, the data is not changed. If there is a failure during one of the successive stages, data can be restored manually. The exception is if the old files were deleted from the file system but the data for the new files did not get written to the disk and was lost.

The ALTER query for changing columns is replicated. The instructions are saved in ZooKeeper, then each replica applies them. All ALTER queries are run in the same order. The query waits for the appropriate actions to be completed on the other replicas. However, a query to change columns in a replicated table can be interrupted, and all actions will be performed asynchronously.

ALTER Query Limitations

The ALTER query lets you create and delete separate elements (columns) in nested data structures, but not whole nested data structures. To add a nested data structure, you can add columns with a name like name.nested_name and the type Array(T). A nested data structure is equivalent to multiple array columns with a name that has the same prefix before the dot.

There is no support for deleting columns in the primary key or the sampling key (columns that are used in the ENGINE expression). Changing the type for columns that are included in the primary key is only possible if this change does not cause the data to be modified (for example, it is allowed to add values to an Enum or to change a type from DateTime to UInt32).

If the ALTER query is not sufficient to make the table changes you need, you can create a new table, copy the data to it using the INSERT SELECT query, then switch the tables using the RENAME query and delete the old table. You can use the clickhouse-copier as an alternative to the INSERT SELECT query.

The ALTER query blocks all reads and writes for the table. In other words, if a long SELECT is running at the time of the ALTER query, the ALTER query will wait for it to complete. At the same time, all new queries to the same table will wait while this ALTER is running.

For tables that don't store data themselves (such as Merge and Distributed), ALTER just changes the table structure, and does not change the structure of subordinate tables. For example, when running ALTER for a Distributed table, you will also need to run ALTER for the tables on all remote servers.

Manipulations With Key Expressions

The following command is supported:

MODIFY ORDER BY new_expression

It only works for tables in the MergeTree family (including replicated tables). The command changes the sorting key of the table to new_expression (an expression or a tuple of expressions). Primary key remains the same.

The command is lightweight in a sense that it only changes metadata. To keep the property that data part rows are ordered by the sorting key expression you cannot add expressions containing existing columns to the sorting key (only columns added by the ADD COLUMN command in the same ALTER query).

Manipulations With Data Skipping Indices

It only works for tables in the *MergeTree family (including replicated tables). The following operations are available:

  • ALTER TABLE [db].name ADD INDEX name expression TYPE type GRANULARITY value AFTER name [AFTER name2] - Adds index description to tables metadata.

  • ALTER TABLE [db].name DROP INDEX name - Removes index description from tables metadata and deletes index files from disk.

These commands are lightweight in a sense that they only change metadata or remove files. Also, they are replicated (syncing indices metadata through ZooKeeper).

Manipulations With Partitions and Parts

The following operations with partitions are available:

DETACH PARTITION

ALTER TABLE table_name DETACH PARTITION partition_expr

Moves all data for the specified partition to the detached directory. The server forgets about the detached data partition as if it does not exist. The server will not know about this data until you make the ATTACH query.

Example:

ALTER TABLE visits DETACH PARTITION 201901

Read about setting the partition expression in a section How to specify the partition expression.

After the query is executed, you can do whatever you want with the data in the detached directory — delete it from the file system, or just leave it.

This query is replicated it moves the data to the detached directory on all replicas. Note that you can execute this query only on a leader replica. To find out if a replica is a leader, perform the SELECT query to the system.replicas table. Alternatively, it is easier to make a DETACH query on all replicas - all the replicas throw an exception, except the leader replica.

DROP PARTITION

ALTER TABLE table_name DROP PARTITION partition_expr

Deletes the specified partition from the table. This query tags the partition as inactive and deletes data completely, approximately in 10 minutes.

Read about setting the partition expression in a section How to specify the partition expression.

The query is replicated it deletes data on all replicas.

ATTACH PARTITION|PART

ALTER TABLE table_name ATTACH PARTITION|PART partition_expr

Adds data to the table from the detached directory. It is possible to add data for an entire partition or for a separate part. Examples:

ALTER TABLE visits ATTACH PARTITION 201901;
ALTER TABLE visits ATTACH PART 201901_2_2_0;

Read more about setting the partition expression in a section How to specify the partition expression.

This query is replicated. Each replica checks whether there is data in the detached directory. If the data is in this directory, the query checks the integrity, verifies that it matches the data on the server that initiated the query. If everything is correct, the query adds data to the replica. If not, it downloads data from the query requestor replica, or from another replica where the data has already been added.

So you can put data to the detached directory on one replica, and use the ALTER ... ATTACH query to add it to the table on all replicas.

REPLACE PARTITION

ALTER TABLE table2 REPLACE PARTITION partition_expr FROM table1

This query copies the data partition from the table1 to table2. Note that data won't be deleted from table1.

For the query to run successfully, the following conditions must be met:

  • Both tables must have the same structure.
  • Both tables must have the same partition key.

CLEAR COLUMN IN PARTITION

ALTER TABLE table_name CLEAR COLUMN column_name IN PARTITION partition_expr

Resets all values in the specified column in a partition. If the DEFAULT clause was determined when creating a table, this query sets the column value to a specified default value.

Example:

ALTER TABLE visits CLEAR COLUMN hour in PARTITION 201902

FREEZE PARTITION

ALTER TABLE table_name FREEZE [PARTITION partition_expr]

This query creates a local backup of a specified partition. If the PARTITION clause is omitted, the query creates the backup of all partitions at once.

Note that for old-styled tables you can specify the prefix of the partition name (for example, '2019') - then the query creates the backup for all the corresponding partitions. Read about setting the partition expression in a section How to specify the partition expression.

!!! note The entire backup process is performed without stopping the server.

At the time of execution, for a data snapshot, the query creates hardlinks to a table data. Hardlinks are placed in the directory /var/lib/clickhouse/shadow/N/..., where:

  • /var/lib/clickhouse/ is the working ClickHouse directory specified in the config.
  • N is the incremental number of the backup.

The same structure of directories is created inside the backup as inside /var/lib/clickhouse/. The query performs 'chmod' for all files, forbidding writing into them.

After creating the backup, you can copy the data from /var/lib/clickhouse/shadow/ to the remote server and then delete it from the local server. Note that the ALTER t FREEZE PARTITION query is not replicated. It creates a local backup only on the local server.

The query creates backup almost instantly (but first it waits for the current queries to the corresponding table to finish running).

ALTER TABLE t FREEZE PARTITION copies only the data, not table metadata. To make a backup of table metadata, copy the file /var/lib/clickhouse/metadata/database/table.sql

To restore data from a backup, do the following:

  1. Create the table if it does not exist. To view the query, use the .sql file (replace ATTACH in it with CREATE).
  2. Copy the data from the data/database/table/ directory inside the backup to the /var/lib/clickhouse/data/database/table/detached/ directory.
  3. Run ALTER TABLE t ATTACH PARTITION queries to add the data to a table.

Restoring from a backup doesn't require stopping the server.

For more information about backups and restoring data, see the Data Backup section.

FETCH PARTITION

ALTER TABLE table_name FETCH PARTITION partition_expr FROM 'path-in-zookeeper'

Downloads a partition from another server. This query only works for the replicated tables.

The query does the following:

  1. Downloads the partition from the specified shard. In 'path-in-zookeeper' you must specify a path to the shard in ZooKeeper.
  2. Then the query puts the downloaded data to the detached directory of the table_name table. Use the ATTACH PARTITION|PART query to add the data to the table.

For example:

ALTER TABLE users FETCH PARTITION 201902 FROM '/clickhouse/tables/01-01/visits';
ALTER TABLE users ATTACH PARTITION 201902;

Note that:

  • The ALTER ... FETCH PARTITION query isn't replicated. It places the partition to the detached directory only on the local server.
  • The ALTER TABLE ... ATTACH query is replicated. It adds the data to all replicas. The data is added to one of the replicas from the detached directory, and to the others - from neighboring replicas.

Before downloading, the system checks if the partition exists and the table structure matches. The most appropriate replica is selected automatically from the healthy replicas.

Although the query is called ALTER TABLE, it does not change the table structure and does not immediately change the data available in the table.

How To Set Partition Expression

You can specify the partition expression in ALTER ... PARTITION queries in different ways:

  • As a value from the partition column of the system.parts table. For example, ALTER TABLE visits DETACH PARTITION 201901.
  • As the expression from the table column. Constants and constant expressions are supported. For example, ALTER TABLE visits DETACH PARTITION toYYYYMM(toDate('2019-01-25')).
  • Using the partition ID. Partition ID is a string identifier of the partition (human-readable, if possible) that is used as the names of partitions in the file system and in ZooKeeper. The partition ID must be specified in the PARTITION ID clause, in a single quotes. For example, ALTER TABLE visits DETACH PARTITION ID '201901'.
  • In the ALTER ATTACH PART query, to specify the name of a part, use a value from the name column of the system.parts table. For example, ALTER TABLE visits ATTACH PART 201901_1_1_0.

Usage of quotes when specifying the partition depends on the type of partition expression. For example, for the String type, you have to specify its name in quotes ('). For the Date and Int* types no quotes are needed.

For old-style tables, you can specify the partition either as a number 201901 or a string '201901'. The syntax for the new-style tables is stricter with types (similar to the parser for the VALUES input format).

All the rules above are also true for the OPTIMIZE query. If you need to specify the only partition when optimizing a non-partitioned table, set the expression PARTITION tuple(). For example:

OPTIMIZE TABLE table_not_partitioned PARTITION tuple() FINAL;

The examples of ALTER ... PARTITION queries are demonstrated in the tests 00502_custom_partitioning_local and 00502_custom_partitioning_replicated_zookeeper.

Synchronicity of ALTER Queries

For non-replicatable tables, all ALTER queries are performed synchronously. For replicatable tables, the query just adds instructions for the appropriate actions to ZooKeeper, and the actions themselves are performed as soon as possible. However, the query can wait for these actions to be completed on all the replicas.

For ALTER ... ATTACH|DETACH|DROP queries, you can use the replication_alter_partitions_sync setting to set up waiting. Possible values: 0 do not wait; 1 only wait for own execution (default); 2 wait for all.

Mutations

Mutations are an ALTER query variant that allows changing or deleting rows in a table. In contrast to standard UPDATE and DELETE queries that are intended for point data changes, mutations are intended for heavy operations that change a lot of rows in a table.

The functionality is in beta stage and is available starting with the 1.1.54388 version. Currently *MergeTree table engines are supported (both replicated and unreplicated).

Existing tables are ready for mutations as-is (no conversion necessary), but after the first mutation is applied to a table, its metadata format becomes incompatible with previous server versions and falling back to a previous version becomes impossible.

Currently available commands:

ALTER TABLE [db.]table DELETE WHERE filter_expr

The filter_expr must be of type UInt8. The query deletes rows in the table for which this expression takes a non-zero value.

ALTER TABLE [db.]table UPDATE column1 = expr1 [, ...] WHERE filter_expr

The command is available starting with the 18.12.14 version. The filter_expr must be of type UInt8. This query updates values of specified columns to the values of corresponding expressions in rows for which the filter_expr takes a non-zero value. Values are casted to the column type using the CAST operator. Updating columns that are used in the calculation of the primary or the partition key is not supported.

One query can contain several commands separated by commas.

For *MergeTree tables mutations execute by rewriting whole data parts. There is no atomicity - parts are substituted for mutated parts as soon as they are ready and a SELECT query that started executing during a mutation will see data from parts that have already been mutated along with data from parts that have not been mutated yet.

Mutations are totally ordered by their creation order and are applied to each part in that order. Mutations are also partially ordered with INSERTs - data that was inserted into the table before the mutation was submitted will be mutated and data that was inserted after that will not be mutated. Note that mutations do not block INSERTs in any way.

A mutation query returns immediately after the mutation entry is added (in case of replicated tables to ZooKeeper, for nonreplicated tables - to the filesystem). The mutation itself executes asynchronously using the system profile settings. To track the progress of mutations you can use the system.mutations table. A mutation that was successfully submitted will continue to execute even if ClickHouse servers are restarted. There is no way to roll back the mutation once it is submitted, but if the mutation is stuck for some reason it can be cancelled with the KILL MUTATION query.

Entries for finished mutations are not deleted right away (the number of preserved entries is determined by the finished_mutations_to_keep storage engine parameter). Older mutation entries are deleted.

Original article