How ClickHouse got fast, how it got here, and where it goes next.
History, courses, tutorials and tools for the column store — how it works, why it was built that way, and how to run it well.
The best of the site
Four guides worth your time first — each one goes deeper than the documentation.
The ClickHouse Production Incident Playbook
Eight classes of production incident — symptom, first checks, mechanism, fix — from real on-call work.
How MergeTree stores data
Parts, marks, merges, mutations and TTL, each one taken apart on a real server.
Materialized views in production
The sharpest tool in ClickHouse and its biggest footgun: semantics, backfills, JOIN traps, refreshable views.
Upgrading ClickHouse safely: from changelog to rollback
What compatibility really pins, what breaks across releases, and a pre-flight checklist with SQL.
The Evolution of ClickHouse
Product, people and AI — three chapters on how a Yandex side project became a database company.
The Product
How ClickHouse went from a reporting prototype inside Yandex.Metrica to an open-source database and then a cloud platform, told by architectural era, with the technical pressure behind each change and what it means for whoever runs it today.
The Brains
The people and the engineering culture behind ClickHouse, how the company came out of Yandex and was funded, and the run of acquisitions that turned a database engine into a set of complete products.
Artificial Intelligence
No chatbot, and no plan for one. The case that ClickHouse's opportunity is infrastructure for agents — where they message, share memory and coordinate — with the SQL that maps onto it and the walls it hits.
Field notes
Blog · Tutorials · Runbooks · News ·Philosophy · Feature requests
- BlogINSERT … RETURNING can return rows ClickHouse never storedOctober 2026
- NewsClickHouse 26.10: function composition, joins on any condition and a long upgrade checklistOctober 2026
- BlogAn object-storage bill that jumped, and the bytes were not the reasonOctober 2026
- BlogDeleting a slice of a huge tableOctober 2026
- BlogWhat REVOKE really revokesOctober 2026
- BlogWhen the system tables ate the diskOctober 2026
Part pressure calculator
How fast does your insert pattern create MergeTree parts, how big are they, and how far is it from TOO_MANY_PARTS? Accounts for async inserts and materialized views.
Talk ClickHouse
Ask questions, share what you built, and suggest what comes next — in the open, on GitHub Discussions.