Snowflake Horizon catalog

ClickHouse can connect to Snowflake Horizon Catalog through the Iceberg REST API that Horizon exposes (powered by Apache Polaris). This lets you read and write Snowflake-managed Iceberg tables from ClickHouse using the DataLakeCatalog database engine with catalog_type = 'horizon'. Horizon is related to, but not the same as, Snowflake Open Catalog / self-hosted Polaris:

	Open Catalog / Polaris (`catalog_type = 'rest'`)	Horizon (`catalog_type = 'horizon'`)
Endpoint	Open Catalog or self-hosted Polaris URI	`https://<org>-<account>.snowflakecomputing.com/polaris/api/catalog`
Warehouse	Polaris catalog / warehouse name	Snowflake database name (usually uppercase)
Auth scope	`PRINCIPAL_ROLE:ALL` (typical)	`session:role:<ROLE>`
Credential	`client_id:client_secret`	PAT or key-pair JWT as `catalog_credential` (OAuth client_secret; not split on ’:’), or bearer `auth_header`

As this feature is beta, enable it with: SET allow_experimental_database_iceberg = 1; (or SET allow_database_iceberg = 1; depending on your ClickHouse version).

Prerequisites

A Snowflake account with Snowflake-managed Iceberg tables
Horizon Iceberg REST endpoint: https://<organization>-<account>.snowflakecomputing.com/polaris/api/catalog
A Snowflake role with privileges on the Iceberg tables (and write privileges if you will INSERT)
Authentication via one of:
- Programmatic Access Token (PAT)
- Key-pair JWT exchanged for an access token
- External OAuth access token (as bearer auth_header)
Object storage reachable from ClickHouse (vended credentials are recommended)
ClickHouse with DataLakeCatalog Iceberg support

Creating a connection

Option A: Programmatic Access Token (recommended)

SET allow_experimental_database_iceberg = 1;

CREATE DATABASE horizon_catalog
ENGINE = DataLakeCatalog('https://<org>-<account>.snowflakecomputing.com/polaris/api/catalog')
SETTINGS
    catalog_type = 'horizon',
    warehouse = 'ICEBERG_TEST_DB',
    catalog_credential = '<PAT>',
    auth_scope = 'session:role:DATA_ENGINEER',
    oauth_server_uri = 'https://<org>-<account>.snowflakecomputing.com/polaris/api/catalog/v1/oauth/tokens',
    vended_credentials = 1;

warehouse must be the Snowflake database name (not a Snowflake virtual warehouse). Unquoted Snowflake identifiers are uppercase.

Option B: Pre-exchanged bearer access token

CREATE DATABASE horizon_catalog
ENGINE = DataLakeCatalog('https://<org>-<account>.snowflakecomputing.com/polaris/api/catalog')
SETTINGS
    catalog_type = 'horizon',
    warehouse = 'ICEBERG_TEST_DB',
    auth_header = 'Authorization: Bearer <ACCESS_TOKEN>',
    vended_credentials = 1;

Query Iceberg tables

USE horizon_catalog;
SHOW TABLES;

SELECT count(*) FROM `PUBLIC.test_table`;
SHOW CREATE TABLE `PUBLIC.test_table`;

Namespace (schema) and table are typically addressed as `SCHEMA.table` unless experimental table namespaces are enabled.

Write path

With a role that has INSERT/UPDATE/DELETE (and CREATE ICEBERG TABLE when creating tables), ClickHouse can write through the same catalog:

-- Insert into an existing Snowflake-managed Iceberg table
INSERT INTO horizon_catalog.`PUBLIC.test_table`
SELECT
    number AS id,
    concat('name_', toString(number)) AS name
FROM numbers(100);

-- Create a new Iceberg table in the Horizon catalog (requires CREATE ICEBERG TABLE)
CREATE TABLE horizon_catalog.`PUBLIC.clickhouse_written`
(
    id Int64,
    name String
)
ENGINE = Iceberg;

Exact CREATE TABLE syntax follows ClickHouse Iceberg / DataLakeCatalog write support for REST catalogs on your version; privilege failures from Snowflake surface as catalog HTTP errors.

Loading into MergeTree

CREATE TABLE my_clickhouse_table
(
    id Int64,
    name String
)
ENGINE = MergeTree
ORDER BY id;

INSERT INTO my_clickhouse_table
SELECT * FROM horizon_catalog.`PUBLIC.test_table`;

ClickHouse guides

Use case guides

Prerequisites

Creating a connection

Option A: Programmatic Access Token (recommended)

Option B: Pre-exchanged bearer access token

Query Iceberg tables

Write path

Loading into MergeTree

​Prerequisites

​Creating a connection

​Option A: Programmatic Access Token (recommended)

​Option B: Pre-exchanged bearer access token

​Query Iceberg tables

​Write path

​Loading into MergeTree

Prerequisites

Creating a connection

Option A: Programmatic Access Token (recommended)

Option B: Pre-exchanged bearer access token

Query Iceberg tables

Write path

Loading into MergeTree