feat: clickhouse_cluster destination

[jorritsandbrink]

Description

This PR adds a new destination clickhouse_cluster that:

• builds on top of clickhouse destination
• adapts generated SQL to work well in cluster setups
• extends functionality to support creation of additional distributed tables

Related Issues

• Closes Cluster support for clickhouse #2200

Additional Context

Related PR: #2573

[cloudflare-workers-and-pages]

Deploying with    Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	docs	1691eac	Commit Preview URL Branch Preview URL	Feb 06 2026, 07:31 AM

[jorritsandbrink]

This bug got solved by adding an insert_deduplication_token to INSERT INTO statements.

[jorritsandbrink]

@rudolfix Could you you review this PR?

I struggled a lot trying to get consistent results, especially with the delete-insert merge strategy, because it relies on subqueries which ClickHouse doesn't seem to like when using replicated/distributed tables. Spent too much time debugging flaky tests, but it seems they're all fixed now.

Notes:

• when working with distributed tables, SELECT and INSERT operations should be done on the distributed table, while other operations (e.g. TRUNCATE, DELETE FROM) should be done on the base table
• when querying a table trough pipeline.dataset(), it will resolve to the distributed table (querying the base table would only return data from one shard)