ClickHouse

A few years back, we introduced ClickHouse Fiddle: a SQL playground where you can test queries without installing anything 💻 It's still one of the fastest ways to experiment with ClickHouse: 🔹 Run queries on any ClickHouse version from your browser 🔹 Create tables, insert data, and execute DDL statements 🔹 Share results with a unique link 🔹 Isolated execution via Docker containers (your data stays private) 🔹 Typical query execution in ~2 seconds The architecture is elegant: for each request, a fresh Docker container spins up with your chosen ClickHouse version, runs your queries, and gets torn down. No always-on instances required. Unlike read-only playgrounds, Fiddle lets you write data and test the full SQL lifecycle—perfect for validating bugs, testing new features, or sharing working examples with colleagues. https://lnkd.in/dQcpgibr

Ready for some major ClickHouse news? November's newsletter is here! 🚀 Big news first: We acquired LibreChat to create the "Agentic Data Stack" - where natural language meets blazing-fast analytics. Query your data in plain English while keeping everything 100% open-source. What else? Glad you asked! 🤓 ⚡ QBit vector search - Choose precision at query time with our new bit-plane column type 📊 24x faster JOINs - ClickHouse 25.10 delivers massive speedups with lazy column replication and bloom filter optimizations 🏗️ 170x log compression - Transform raw Nginx logs into structured, queryable data 🔧 AI-powered warehouse - Viralo built their entire data stack from 0→1 using Claude Sonnet and ClickHouse Cloud Plus deep dives on streaming insert monitoring, end-to-end observability with OpenTelemetry traces, and wiring ClickHouse directly into Salesforce. 🙏 Thanks to this month's contributors: William ATTACHE, Raufs Dunamalijevs, Shubham Bhardwaj, Pranav Mehta, Alexey Milovidov, Nisarg Pipaliy, Julian V., Lionel Palacin, Andrii K, Kiran Raparti, and Pragnesh Bhavsar And our featured community member this month is Kiyose Ryu from SmartNews!

Great to see Binance using ClickHouse

48 hours. Streaming data. Real‑time AI agents. We loved sponsoring the AWS MCP Hackathon in San Francisco and seeing what builders shipped with ClickHouse. Three standouts that turned live signals into action: 🥇 Ad Optimizer Agent: Predicts ad bidding prices from time‑series signals. Confluent Kafka streams into ClickHouse for instant feature computation and smarter updates. 🥈 AI Ops Agent: An AI incident commander that stays reliable when APIs or models fail. Temporal handles retries and fallbacks. AWS Bedrock analyzes incidents. Slack delivers clear action plans. ClickHouse stores timelines and metrics for fast root‑cause analysis. 🔎 GlucoTrack Predictive Agent: Monitors and forecasts glucose trends using the OhioT1DM dataset. ClickHouse powers sub‑second queries so dashboards feel like a real product, not a demo. Common pattern across teams: streaming events into ClickHouse, computing features in real time, and letting agents act immediately. Builders paired this with Temporal for orchestration and Bedrock for reasoning. The result: instant analytics in production‑style UIs. https://lnkd.in/e_UdpZBU

Welcome back to House Party. This time, it's the SQL. 🕺 🪩 We’re bringing the Chainsmokers back to Vegas for another epic night of music, food, and drink. If you’ll be in town for AWS re:Invent, you won’t want to miss it. 👉 RSVP at https://lnkd.in/dvy7D7HP

When Idea Clan’s new all-in-one marketing platform FabFunnel outgrew MySQL, the team made a shortlist of what they were looking for: ⚡️ Faster inserts and real-time analytics 📊 A columnar database for denormalized data 🧠 Familiar SQL syntax for an easy migration ☁️ Minimal operational overhead At a ClickHouse meetup in Delhi, engineers Anmol Jain and Sidhant Gaba shared how those requirements led them to ClickHouse Cloud, and how it reshaped their reporting stack: ✅ Queries that once took 11s now return in <1s ✅ 19.2M queries processed in a single month ✅ 11.67T rows read, 606 TB of data handled ✅ Predictable costs while scaling to billions of rows Read more on our blog 👇 https://lnkd.in/eiQZWn5i

🎤 Bangkok, you’re up! ICYMI, we have Open Mic Lightning Talks at the ClickHouse Bangkok Meetup next Tuesday, November 25. Got a spicy query, a tiny tip, or a 3–5 min ClickHouse story? Jump on the mic! 🐣 First-time speakers welcome. 🦁 Brave volunteers get a limited-edition swag pack (black hoodie, tumbler, mini bluetooth speaker) — while they last! 📍 AWS Office @ Singha Complex Building 🗓️ Tue, Nov 25 ⏰ Registration opens at 6:30 PM ✅ RSVP: https://lnkd.in/dbNemzVA Come learn something fast, meet great people, and leave cooler than you arrived. See you there!

⚠️ Mixing memory allocators is asking for trouble. Upgrading chDB to ClickHouse 25.8 exposed exactly this problem: when you embed ClickHouse in Python, you're mixing two different allocators (Python's malloc and jemalloc). If memory allocated by one gets freed by the other, it crashes. 💥 Our solution: runtime memory fingerprinting using je_mallctl("arenas.lookup") to detect which allocator owns each pointer. But arenas.lookup itself would crash on invalid pointers. Auxten Wang fixed this upstream in jemalloc by adding boundary checking, which allows it to handle arbitrary pointers safely. Patch merged, he's now an official jemalloc contributor. 🎉 Bonus: Adding thread-local checks to skip the lookup in hot paths resulted in a 61x speedup on one benchmark query. 🚀 https://lnkd.in/dePp8JRB

A few years ago, Klaviyo hit a scaling wall. Their segmentation engine took over an hour to evaluate customer segments, with logic spread across Python, MySQL, and Cassandra. For a rapidly growing marketing platform, they needed something faster. They rebuilt the whole thing on ClickHouse. Now what used to take an hour happens in one second. 🔸 192-node cluster handling billions of daily updates 🔸 Smart sharding by company then customer profile 🔸 Only recomputes segments that actually changed 🔸 Processes tens of billions of membership changes in near real-time Mark Needham breaks down how they pulled it off, based on an interview with Patrick McGrath https://lnkd.in/etpMxtQs