Support matrix

Comprehensive support matrices for ClickHouse open table format integrations and data catalog connections.

This page provides comprehensive support matrices for ClickHouse’s data lake integrations. It covers the features available for each open table format, the catalogs ClickHouse can connect to, and the capabilities supported by each catalog.

Open table format support

ClickHouse integrates with four open table formats: Apache Iceberg, Delta Lake, Apache Hudi, and Apache Paimon. Select a format below to view its support matrix.

Legend: ✅ Supported | ⚠️ Partial / Experimental | ❌ Not supported

Apache Iceberg

Feature	Status	Notes
Storage backends
AWS S3	✅	Via icebergS3() or iceberg() alias
GCS	✅	Via icebergS3() or iceberg() alias
Azure Blob Storage	✅	Via icebergAzure()
HDFS	⚠️	Via icebergHDFS(). Deprecated.
Local filesystem	✅	Via icebergLocal()
Access methods
Table function	✅	icebergS3() with variants per backend
Table engine	✅	IcebergS3 with variants per backend
Cluster-distributed reads	✅	icebergS3Cluster, icebergAzureCluster, icebergHDFSCluster
Named collections	✅	Defining a named collection
Read features
Read support	✅	Full SELECT support with all ClickHouse SQL functions
Partition pruning	✅	See Partition pruning.
Hidden partitioning	✅	Iceberg transform-based partitioning supported
Partition evolution	✅	Reading tables with changing partition specs over time supported
Schema evolution	✅	Column addition, removal, and reordering. See Schema evolution.
Type promotion / widening	✅	int → long, float → double, decimal(P,S) → decimal(P',S) where P’ > P. See Schema evolution.
Time travel / snapshots	✅	Via iceberg_timestamp_ms or iceberg_snapshot_id settings. See Time travel.
Position deletes	✅	See Processing deleted rows.
Equality deletes	✅	Table engine only, from v25.8+. See Processing deleted rows.
Merge-on-read	⚠️	Experimental. Supported for delete operations.
Format versions	⚠️	v1 and v2 supported. V3 not supported.
Column statistics	✅
Bloom filters / puffin files	❌	Bloom filter indexes in puffin files not supported
Virtual columns	✅	_path, _file, _size, _time, _etag. See Virtual columns.
Write features
Table creation	✅	Experimental. Requires allow_insert_into_iceberg = 1. From v25.7+. See Creating a table.
INSERT	✅	Beta from 26.2. Requires allow_insert_into_iceberg = 1. See Inserting data.
DELETE	✅	Experimental. Requires allow_insert_into_iceberg = 1. Via ALTER TABLE ... DELETE WHERE. See Deleting data.
ALTER TABLE (schema changes)	✅	Experimental. Requires allow_insert_into_iceberg = 1. Add, drop, modify, rename columns. See Schema evolution.
Compaction	⚠️	Experimental. Requires allow_experimental_iceberg_compaction = 1. Merges position delete files into data files. See Compaction. Other Iceberg compaction operations not supported.
UPDATE / MERGE	❌	Not supported. See Compaction.
Copy-on-write	❌	Not supported
Expire snapshots	❌	Not supported
Remove orphan files	❌	Not supported
Writing partitions	✅	Supported.
Altering partitions	❌	The changing of the partitioning scheme from ClickHouse is not supported. ClickHouse can write to iceberg tables which have an evolved partitioning.
Metadata
Branching and tagging	❌	Iceberg branch/tag references not supported
Metadata file resolution	✅	Support for metadata resolution through catalogs, simple directory listing, using ‘version-hint’ and specific path. Configurable via iceberg_metadata_file_path and iceberg_metadata_table_uuid. See Metadata file resolution.
Data caching	✅	Same mechanism as S3/Azure/HDFS storage engines. See Data cache.
Metadata caching	✅	Manifest and metadata files cached in memory. Enabled by default via use_iceberg_metadata_files_cache. See Metadata cache.

Delta Lake

From version 25.6, ClickHouse reads Delta Lake tables using the Delta Lake Rust kernel, providing broader feature support; however, known issues occur when accessing data in Azure Blob Storage. For this reason the Kernel is disabled when reading data on Azure Blob Storage. We indicate below which features require this kernel.

Feature	Status	Notes
Storage backends
AWS S3	✅	Via deltaLake() or deltaLakeS3()
GCS	✅	Via deltaLake() or deltaLakeS3()
Azure Blob Storage	✅	Via deltaLakeAzure()
HDFS	❌	Not supported
Local filesystem	✅	Via deltaLakeLocal()
Access methods
Table function	✅	deltaLake() with variants per backend
Table engine	✅	DeltaLake
Cluster-distributed reads	✅	deltaLakeCluster, deltaLakeAzureCluster
Named collections	✅	Named collection
Read features
Read support	✅	Full SELECT support with all ClickHouse SQL functions
Partition pruning	✅	Requires Delta Kernel.
Schema evolution	✅	Requires Delta Kernel.
Time travel	✅	Requires Delta Kernel.
Deletion vectors	✅
Column mapping	✅
Change data feed	✅	Requires Delta Kernel.
Virtual columns	✅	_path, _file, _size, _time, _etag. See Virtual columns.
Write features
INSERT	✅	Experimental. Requires allow_experimental_delta_lake_writes = 1. See DeltaLake engine. Requires Delta Kernel.
DELETE / UPDATE / MERGE	❌	Not supported
CREATE empty table	❌	Creation of a new empty Delta Lake table is not supported. CREATE TABLE operation assumes existence of existing Delta Lake on object storage.
Caching
Data caching	✅	Same mechanism as S3/Azure/HDFS storage engines. See Data cache.

Apache Hudi

Feature	Status	Notes
Storage backends
AWS S3	✅	Via hudi()
GCS	✅	Via hudi()
Azure Blob Storage	❌	Not supported
HDFS	❌	Not supported
Local filesystem	❌	Not supported
Access methods
Table function	✅	hudi()
Table engine	✅	Hudi
Cluster-distributed reads	✅	hudiCluster (S3 only)
Named collections	✅	Hudi arguments
Read features
Read support	✅	Full SELECT support with all ClickHouse SQL functions
Schema evolution	❌	Not supported
Time travel	❌	Not supported
Virtual columns	✅	_path, _file, _size, _time, _etag. See Virtual columns.
Write features
INSERT / DELETE / UPDATE	❌	Read-only integration
Caching
Data caching	❌	Not supported

Apache Paimon

Feature	Status	Notes
Storage backends
S3	✅	Experimental. Via paimon() or paimonS3()
GCS	✅	Experimental. Via paimon() or paimonS3()
Azure Blob Storage	✅	Experimental. Via paimonAzure()
HDFS	⚠️	Experimental. Via paimonHDFS(). Deprecated.
Local filesystem	✅	Experimental. Via paimonLocal()
Access methods
Table function	✅	Experimental. paimon() with variants per backend
Table engine	❌	No dedicated table engine
Cluster-distributed reads	✅	Experimental. paimonS3Cluster, paimonAzureCluster, paimonHDFSCluster
Named collections	✅	Experimental. Defining a named collection
Read features
Read support	✅	Experimental. Full SELECT support with all ClickHouse SQL functions
Schema evolution	❌	Not supported
Time travel	❌	Not supported
Virtual columns	✅	Experimental. _path, _file, _size, _time, _etag. See Virtual columns.
Write features
INSERT / DELETE / UPDATE	❌	Read-only integration
Caching
Data caching	❌	Not supported

Catalog support

ClickHouse can connect to external data catalogs using the DataLakeCatalog database engine, which exposes the catalog as a ClickHouse database. Tables registered in the catalog appear automatically and can be queried with standard SQL.

The following catalogs are currently supported. Refer to each catalog’s reference guide for full setup instructions.

Catalog	Formats	Read	Create table	INSERT	Reference guide
AWS Glue Catalog	Iceberg	✅ Beta	❌	❌	Glue catalog guide
BigLake Metastore	Iceberg	✅ Beta	❌	❌	BigLake Metastore guide
Databricks Unity Catalog	Delta, Iceberg	✅ Beta	✅ Beta	✅ Beta	Unity Catalog guide
Iceberg REST	Iceberg	✅ Beta	❌	❌	REST catalog guide
Lakekeeper	Iceberg	✅ Beta	❌	❌	Lakekeeper catalog guide
Project Nessie	Iceberg	✅ Experimental	❌	❌	Nessie catalog guide
Microsoft OneLake	Iceberg	✅ Beta	✅ Beta	✅ Beta	OneLake catalog guide

All catalog integrations currently require an experimental or beta setting to be enabled. With the exception of Microsoft OneLake and Databricks Unity Catalog, all catalogs expose read-only access — tables can be queried but not created or written to through the catalog connection. To load data from a catalog into ClickHouse for faster analytics, use INSERT INTO SELECT as described in the accelerating analytics guide. To write data back to open table formats, create standalone Iceberg tables as described in the writing data guide.