August 2026 newsletter

wpnews.pro

A new database research lab, an interactive playground for over 100 databases, and our biggest release post ever - it’s just your average month in the world of ClickHouse!

In this issue, meet Andy Pavlo, the newest member of the ClickHouse team, learn about what’s new in 26.7, and catch up on the latest in ClickHouse Managed Postgres, observability, vector search, and more.

And don’t forget to register for the Open House Roadshow, which is visiting six cities over the next six weeks.

Featured community member: Manuel Raimann #

This month's featured community member is Manuel Raimann, Head of Engineering at Redact.dev.

Manuel contributed a series of performance improvements in the recent ClickHouse 26.7 release. His work sped up wide-integer comparisons by up to 7×, Delta decompression by up to 5×, and statistical and bitwise aggregates by up to 4×. He also made numeric array operations and primary-key index filtering faster.

Apart from contributing to ClickHouse, he also contributes to the open-source Deadlock API, which provides game data and analytics for Deadlock, Valve’s upcoming game combining third-person shooting with MOBA-style gameplay.

➡️ Connect with Manuel on LinkedIn

Open House Roadshow #

The Open House Roadshow continues, with visits to Amsterdam (Sep 1), Singapore (Sep 24), and London (Sep 30).

Each stop will feature talks from both ClickHouse employees and customers, as well as optional training workshops.

We’ll also be visiting New York City, Bangalore, and Munich.

➡️ See all Open House locations

26.7 release #

The 26.7 release blog post is our longest one yet. While there is certainly value in brevity, we’re going to let ourselves off this time because there were so many features that we needed to tell you about!

As always, we have join optimizations. The hash join algorithm has been made more efficient by reducing the probe-side input and shrinking build-side hash tables. ClickHouse will now also automatically choose how to arrange joins in a multi-table query.

Text indexes now store token positions, a feature that’s needed for efficient phrase search. We tested it on a Hacker News dataset, and a query for the phrase "Google web search"

was 40 times faster.

The EXPLAIN clause has a new member: EXPLAIN ANALYZE

, which runs every query-processing phase before discarding the result rows and annotating the logical execution-plan tree with measurements collected during actual execution.

That’s just a brief taster - we’ll let you read the blog post to find out the rest!

Andy Pavlo joins ClickHouse to establish ClickHouse Labs #

The news of the month (and maybe the year?!) is that Andy Pavlo is joining ClickHouse to establish ClickHouse Labs. Andy was previously a professor in the Computer Science Department at Carnegie Mellon University since 2013.

ClickHouse Labs is an applied research lab that will conduct scientifically valuable research and then help transform the best ideas into technology that matters to users.

In addition to developing ideas to improve ClickHouse, the lab will work with the ClickHouse PostgreSQL team to help establish its managed service as a market leader in performance and reliability.

Your TTL is working. That might be the problem. #

Amrelboridy’s disk kept filling up, but there wasn’t any indication why. There weren’t any orphaned or detached parts, and data retention was healthy.

In this blog post, Amrelboridy explains what happens in ClickHouse when a row’s Time To Live (TTL) expires and how configuring the partition key can save you a lot of headaches.

Alexey created a playground for 110 database systems #

Alexey has extended ClickBench, a benchmark for analytical databases, to include an interactive playground where you can run ad hoc queries on the 100m-record ClickBench dataset against over 100 database systems.

The playground runs on Firecracker microVMs on an AWS EC2 instance, and in the blog post, Alexey explains why this was the only approach that provided strong isolation, reasonable cost, and fast startup.

Postgres vs ClickHouse is the wrong question. I use both. #

Artem Senenko explains that “Postgres vs ClickHouse” is the wrong question because the databases solve different problems.

He uses Postgres for mutable data, such as monitors, incidents, teams, and configuration. But the continuously growing stream of immutable check results goes into ClickHouse for efficient storage and fast analytical queries.

The majority of the blog post focuses on ways to optimize ClickHouse, including compression codecs, tenant-aware sort keys, per-row retention periods, and materialized views that pre-aggregate data for dashboards.

What's new in ClickHouse Managed Postgres: Customer notifications, better observability, faster backups, extensions, and more #

In the latest issue of What’s New in ClickHouse Managed Postgres, we introduce a smoother onboarding experience, proactive storage notifications, richer observability, and faster, more predictable backups for databases larger than 10 TB.

We’ve also expanded the extension ecosystem: pg_re2 makes regular-expression queries up to 9× faster, while improvements to pg_clickhouse and pg_stat_ch enable more efficient analytics and query monitoring across Postgres and ClickHouse.

How we build and evaluate our MCP server for SRE agents #

Brandon Pereira explains how the ClickStack team built hdx-evals, an open-source framework for measuring the reliability of AI agents' investigations into production incidents.

Across five reproducible scenarios containing millions of synthetic logs and spans, the purpose-built ClickStack MCP server scored 7–20 percentage points higher than giving agents direct SQL access through the ClickHouse MCP server.

Brandon also explores how tool schemas, response design, actionable errors, and query speed influence agent accuracy and consistency.

Quick reads #

Kaushik Iska explains how ClickHouse Managed Postgres uses tiered, WAL-aware write throttlingto slow clients before an archiving backlog fills the disk, giving the archiver time to recover before automatically restoring full throughput. - Arsad Tanzim builds a ClickHouse, OpenTelemetry, and Grafana prototypeto unify logs, metrics, and traces in one SQL-queryable backend. - Mohamed Hussain S tests ClickHouse vector searchand finds it can replace a dedicated vector database when embeddings sit alongside analytical data, while vector-only workloads with constant high-frequency writes favor a purpose-built system. - Jordan Simonovski turns a Gaggia espresso machine into an OpenTelemetry-instrumented system, streaming ESP32 and household sensor data into ClickStack and using an LLM agent to turn each shot into actionable brewing advice.

Get started today

Interested in seeing how ClickHouse works on your data? Get started with ClickHouse Cloud in minutes and receive $300 in free credits.

Sign up

Upcoming events #

source & further reading

clickhouse.com — original article ClickStack and Hud bring runtime intelligence to AI-powered development How iFood built its agentic security platform on ClickHouse Cloud ClickHouse launches ClickHouse Labs with Andy Pavlo as VP of Database Research

August 2026 newsletter

Featured community member: Manuel Raimann #

Open House Roadshow #

26.7 release #

Andy Pavlo joins ClickHouse to establish ClickHouse Labs #

Your TTL is working. That might be the problem. #

Alexey created a playground for 110 database systems #

Postgres vs ClickHouse is the wrong question. I use both. #

What's new in ClickHouse Managed Postgres: Customer notifications, better observability, faster backups, extensions, and more #

How we build and evaluate our MCP server for SRE agents #

Quick reads #

Get started today

Upcoming events #

Global virtual events

Virtual training

Events in AMER

Events in EMEA

Events in APAC

Run your AI side-project on zahid.host