DB::Exception: Cannot alter column 'col' because it is used in the primary key

Attempting to Modify a Primary Key Column

When executing ALTER TABLE t MODIFY COLUMN col TYPE String on a MergeTree‑family table that already defines col in its primary key, ClickHouse throws the error DB::Exception: Cannot alter column 'col' because it is used in the primary key. The operation is rejected before any data mutation occurs.

Constraints and Uncertainty

• Primary key columns determine the data layout of MergeTree parts; changing them would require rewriting all parts.
• Current ClickHouse releases (up to 23.x) do not support in‑place primary key mutation.
• Documentation lists this as a limitation, but no version has removed it.

The absence of a clear path to alter primary key columns raises a design question: should ClickHouse allow primary key changes through ALTER TABLE with automatic data reorganization, or is recreating the table the only viable approach?

Unresolved Questions

1. What would be the implications of adding support for primary key mutation in MergeTree tables, in terms of performance, consistency, and storage requirements?
2. Are there any experimental or beta releases that already implement a safe, incremental rewrite of parts when a primary key column is altered?
3. Could a hybrid approach—such as a two‑phase migration that first reindexes new parts and then drops old ones—be integrated into the existing ALTER TABLE framework?