Question 1

How well do ClickHouse types survive into DuckDB?

Accepted Answer

Well, because both engines take nested data seriously. Maps map onto DuckDB MAP, named tuples become STRUCT, Nested columns become arrays of structs, and typed arrays keep their element type rather than flattening into JSON. The types worth validating after a first load are the very wide numerics, meaning large unsigned and signed integers and high precision decimals, along with Variant and Dynamic columns, since those can narrow or arrive as strings.

Question 2

Does incremental syncing catch deletes?

Accepted Answer

No. Reading from ClickHouse uses a tracking column rather than a change stream, which is expected for an analytical store, so each run moves rows whose tracking value has advanced. A row deleted at the source stops appearing rather than being flagged, and a row edited without the tracking value moving is missed. Schedule a periodic full resync into MotherDuck if parity matters. With a primary key present, ordinary incremental runs do merge rather than append.

Question 3

Does the ClickHouse table engine matter?

Accepted Answer

It does when ClickHouse is the destination rather than the source. Tables created for keyed data use ReplacingMergeTree ordered on the sync key so upserts deduplicate correctly. For existing tables, ReplacingMergeTree variants accept all write modes while plain MergeTree variants accept append and replace only, and aggregating engines are refused because a retried batch would change the meaning of the aggregate rather than being idempotent.

Question 4

Why would I run both of these together?

Accepted Answer

Usually because they serve different audiences. ClickHouse handles high volume event data and powers product facing analytics, while MotherDuck suits interactive exploration and joining that data to smaller business datasets without provisioning cluster capacity for it. Replicating a curated subset across means analysts work in DuckDB against the tables they need, and the ClickHouse cluster keeps serving the workload it was sized for.

Connect MotherDuck with ClickHouse

Connect MotherDuck and ClickHouse with Polytomic

Structured types preserved

Arrays stay typed

Watermark incremental

Trusted by 1,000+ businesses worldwide

Step-by-step integration guide

Connect MotherDuck in Polytomic

Connect ClickHouse in Polytomic

Create a sync between MotherDuck and ClickHouse

Map fields and enable the sync

Frequently asked questions

The sync and ETL platform that does it all

Powerful data sync

Advanced analytics

Enterprise security

Easy setup

Ready to connect MotherDuck with ClickHouse?