Clone a table on Databricks

You can create a copy of an existing Delta Lake table on Databricks at a specific version using the clone command. Clones can be either deep or shallow.

Databricks also supports cloning Parquet and Iceberg tables. See Incrementally clone Parquet and Iceberg tables to Delta Lake.

For details on using clone with Unity Catalog, see Shallow clone for Unity Catalog tables.

Note

Databricks recommends using Delta Sharing to provide read-only access to tables across different organizations. See What is Delta Sharing?.

Clone types

  • A deep clone is a clone that copies the source table data to the clone target in addition to the metadata of the existing table. Additionally, stream metadata is also cloned such that a stream that writes to the Delta table can be stopped on a source table and continued on the target of a clone from where it left off.

  • A shallow clone is a clone that does not copy the data files to the clone target. The table metadata is equivalent to the source. These clones are cheaper to create.

The metadata that is cloned includes: schema, partitioning information, invariants, nullability. For deep clones only, stream and COPY INTO metadata are also cloned. Metadata not cloned are the table description and user-defined commit metadata.

What are the semantics of Delta clone operations?

If you have working with a Delta table registered to the Hive metastore or a collection of files not registered as a table, clone has the following semantics:

Important

In Databricks Runtime 13.3 LTS and above, Unity Catalog managed tables have support for shallow clones. Clone semantics for Unity Catalog tables differ significantly from Delta Lake clone semantics in other environments. See Shallow clone for Unity Catalog tables.

  • Any changes made to either deep or shallow clones affect only the clones themselves and not the source table.

  • Shallow clones reference data files in the source directory. If you run vacuum on the source table, clients can no longer read the referenced data files and a FileNotFoundException is thrown. In this case, running clone with replace over the shallow clone repairs the clone. If this occurs often, consider using a deep clone instead which does not depend on the source table.

  • Deep clones do not depend on the source from which they were cloned, but are expensive to create because a deep clone copies the data as well as the metadata.

  • Cloning with replace to a target that already has a table at that path creates a Delta log if one does not exist at that path. You can clean up any existing data by running vacuum.

  • For existing Delta tables, a new commit is created that includes the new metadata and new data from the source table. This new commit is incremental, meaning that only new changes since the last clone are committed to the table.

  • Cloning a table is not the same as Create Table As Select or CTAS. A clone copies the metadata of the source table in addition to the data. Cloning also has simpler syntax: you don’t need to specify partitioning, format, invariants, nullability and so on as they are taken from the source table.

  • A cloned table has an independent history from its source table. Time travel queries on a cloned table do not work with the same inputs as they work on its source table.

Example clone syntax

The following code examples demonstrate syntax for creating deep and shallow clones:

CREATE TABLE target_table CLONE source_table; -- Create a deep clone of source_table as target_table

CREATE OR REPLACE TABLE target_table CLONE source_table; -- Replace the target

CREATE TABLE IF NOT EXISTS target_table CLONE source_table; -- No-op if the target table exists

CREATE TABLE target_table SHALLOW CLONE source_table;

CREATE TABLE target_table SHALLOW CLONE source_table VERSION AS OF version;

CREATE TABLE target_table SHALLOW CLONE source_table TIMESTAMP AS OF timestamp_expression; -- timestamp can be like “2019-01-01” or like date_sub(current_date(), 1)

from delta.tables import *

deltaTable = DeltaTable.forName(spark, "source_table")

deltaTable.clone(target="target_table", isShallow=True, replace=False) # clone the source at latest version

deltaTable.cloneAtVersion(version=1, target="target_table", isShallow=True, replace=False) # clone the source at a specific version

# clone the source at a specific timestamp such as timestamp="2019-01-01"
deltaTable.cloneAtTimestamp(timestamp="2019-01-01", target="target_table", isShallow=True, replace=False)

import io.delta.tables._

val deltaTable = DeltaTable.forName(spark, "source_table")

deltaTable.clone(target="target_table", isShallow=true, replace=false) // clone the source at latest version

deltaTable.cloneAtVersion(version=1, target="target_table", isShallow=true, replace=false) // clone the source at a specific version

deltaTable.cloneAtTimestamp(timestamp="2019-01-01", target="target_table", isShallow=true, replace=false) // clone the source at a specific timestamp

For syntax details, see CREATE TABLE CLONE.

Clone metrics

CLONE reports the following metrics as a single row DataFrame once the operation is complete:

  • source_table_size: Size of the source table that’s being cloned in bytes.

  • source_num_of_files: The number of files in the source table.

  • num_removed_files: If the table is being replaced, how many files are removed from the current table.

  • num_copied_files: Number of files that were copied from the source (0 for shallow clones).

  • removed_files_size: Size in bytes of the files that are being removed from the current table.

  • copied_files_size: Size in bytes of the files copied to the table.

Clone metrics example

Permissions

You must configure permissions for Databricks table access control and your cloud provider.

Table access control

The following permissions are required for both deep and shallow clones:

  • SELECT permission on the source table.

  • If you are using CLONE to create a new table, CREATE permission on the database in which you are creating the table.

  • If you are using CLONE to replace a table, you must have MODIFY permission on the table.

Cloud provider permissions

If you have created a deep clone, any user that reads the deep clone must have read access to the clone’s directory. To make changes to the clone, users must have write access to the clone’s directory.

If you have created a shallow clone, any user that reads the shallow clone needs permission to read the files in the original table, since the data files remain in the source table with shallow clones, as well as the clone’s directory. To make changes to the clone, users will need write access to the clone’s directory.

Use clone for data archiving

You can use deep clone to preserve the state of a table at a certain point in time for archival purposes. You can sync deep clones incrementally to maintain an updated state of a source table for disaster recovery.

-- Every month run
CREATE OR REPLACE TABLE archive_table CLONE my_prod_table

Use clone for ML model reproduction

When doing machine learning, you may want to archive a certain version of a table on which you trained an ML model. Future models can be tested using this archived data set.

-- Trained model on version 15 of Delta table
CREATE TABLE model_dataset CLONE entire_dataset VERSION AS OF 15

Use clone for short-term experiments on a production table

To test a workflow on a production table without corrupting the table, you can easily create a shallow clone. This allows you to run arbitrary workflows on the cloned table that contains all the production data but does not affect any production workloads.

-- Perform shallow clone
CREATE OR REPLACE TABLE my_test SHALLOW CLONE my_prod_table;

UPDATE my_test WHERE user_id is null SET invalid=true;
-- Run a bunch of validations. Once happy:

-- This should leverage the update information in the clone to prune to only
-- changed files in the clone if possible
MERGE INTO my_prod_table
USING my_test
ON my_test.user_id <=> my_prod_table.user_id
WHEN MATCHED AND my_test.user_id is null THEN UPDATE *;

DROP TABLE my_test;

Use clone to override table properties

Table property overrides are particularly useful for:

  • Annotating tables with owner or user information when sharing data with different business units.

  • Archiving Delta tables and table history or time travel is required. You can specify the data and log retention periods independently for the archive table. For example:

CREATE OR REPLACE TABLE archive_table CLONE prod.my_table
TBLPROPERTIES (
delta.logRetentionDuration = '3650 days',
delta.deletedFileRetentionDuration = '3650 days'
)

dt = DeltaTable.forName(spark, "prod.my_table")
tblProps = {
"delta.logRetentionDuration": "3650 days",
"delta.deletedFileRetentionDuration": "3650 days"
}
dt.clone(target="archive_table", isShallow=False, replace=True, tblProps)

val dt = DeltaTable.forName(spark, "prod.my_table")
val tblProps = Map(
"delta.logRetentionDuration" -> "3650 days",
"delta.deletedFileRetentionDuration" -> "3650 days"
)
dt.clone(target="archive_table", isShallow = false, replace = true, properties = tblProps)