ClickHouse continues its rapid monthly release cycle, and the 25.8 release (August 28, 2025) is packed with developer-friendly features, cloud/lakehouse improvements, and powerful optimizations. This release also stabilizes long-awaited experimental features, making ClickHouse more reliable in production.
Here’s a breakdown of what’s new and why it matters.
Backward-Incompatible Change
Before upgrading, note these key changes:
- More accurate memory tracking: All external library allocations are now tracked, which may increase reported memory usage or trigger stricter
MEMORY_LIMIT_EXCEEDEDerrors. - JSON handling: Arrays with mixed types now infer as
Array(Dynamic)instead of unnamedTuple. - MergeTree performance boost:
write_marks_for_substreams_in_compact_partsenabled by default, improving subcolumn reads but breaking compatibility with versions <25.5. - Query scheduling fairness: Default scheduler switched to
fair_round_robinfor better handling of single-threaded workloads. - Permissions tightening: Azure, Iceberg, and Delta functions now validate storage permissions more strictly.
- Heredoc syntax: PostgreSQL-style heredoc tags are now restricted to word characters, reducing ambiguity.
New Features
1. PromQL Dialect Support
ClickHouse now supports PromQL queries directly in the client. Analysts can run queries like:
SET dialect='promql';
SELECT * FROM prometheusQuery('up', ...);Currently, functions rate, delta, and increase are supported – bridging time-series observability with ClickHouse analytics.
2. AI-Powered SQL Generation
ClickHouse’s AI-assisted query generation now works out-of-the-box if you provide OPENAI_API_KEY or ANTHROPIC_API_KEY. Zero-config AI SQL is here.
3. ArrowFlight RPC
A new ArrowFlight RPC protocol improves interoperability with the Arrow ecosystem, adding the arrowflight table function for faster data exchange.
4. Iceberg & Delta Lake Enhancements
ClickHouse doubles down on lakehouse support:
- Position and equality deletes for Iceberg
- Writes (create, alter, add/drop/modify columns)
- REST and Glue catalog support
- Stats in metadata for better pruning
- Delta Lake writes and snapshot version reads
- Merge position deletes into data files for efficiency
5. Vector Search Upgrades
Vector similarity index now supports binary quantization, drastically cutting memory and boosting indexing speed. Approximate vector search is now GA.
6. Dead Letter Queues
A new system table captures erroneous messages from streaming engines like Kafka – making debugging ingestion pipelines far easier.
7. More Integrations
- PostgreSQL protocol now supports
COPY - C# client support for MySQL protocol
- Hive partition style reads/writes
- ZooKeeper connection history (
zookeeper_connection_log) - Time-series functions for linear prediction and derivation
- New
Hashoutput format for data “fingerprints”
From Experimental to Stable
Several long-awaited features are promoted:
- Correlated subqueries → now default
- Lightweight updates & deletes → promoted to beta
- Data lake catalogs (Unity, Glue, Rest, Hive) → beta
- Approximate vector search → now GA
Performance Improvements
ClickHouse 25.8 makes queries faster with:
- New Parquet reader (v3) with filter pushdown
- Faster Azure client (ditching the old curl-based library)
- Smarter join optimization & block size memory limits
- Bloom filter index improvements (IN-like efficiency)
- DISTINCT window aggregates now run in linear time
- Better parallel replica workload distribution with rendezvous hashing
- Reduced latency for vector search queries
- Aggregations improved with out-of-order bucket merges
Improvements & Fixes
- Web UI polish: resizable columns, better dashboards, improved dark mode
- Replicated DBs: configurable defaults, faster recovery
- Security: forced secure connections on MySQL/Postgres ports, masked credentials in Iceberg/Delta
- Bug fixes: Numerous stability fixes across Iceberg, Delta, Keeper, lightweight updates, backups, JSON parsing, and distributed queries
Why This Release Matters
- Analysts → PromQL + AI SQL make querying faster and more intuitive.
- Data engineers → Stronger Iceberg/Delta support, ArrowFlight RPC, and lakehouse catalog integrations.
- Ops teams → Better memory tracking, fairer scheduling, and robust logging.
ClickHouse 25.8 strengthens its role not just as a fast OLAP database, but also as a key player in modern lakehouse and observability pipelines.
Upgrade Notes
- Validate memory-heavy queries, as tracking is stricter now.
- Ensure cluster versions >=25.5 if using Compact parts.
- Re-check permissions for Azure/Iceberg/Delta integrations.
- Test Iceberg and Delta workflows in staging before production.
Closing Thoughts
ClickHouse 25.8 brings a well-rounded mix: smarter SQL with PromQL and AI, stronger interoperability with ArrowFlight and lakehouses, and key performance boosts.
If you run ClickHouse in production, this is a release worth upgrading to – especially if you rely on BI dashboards, observability pipelines, or data lakes.
Looking for Expert ClickHouse Solutions?
At Quantrail Data, we offer:
- Fully managed ClickHouse services
- Seamless migration support
- Performance optimization and consulting
Whether you’re deploying ClickHouse at scale, integrating geospatial or lakehouse pipelines, or just want expert backup – we’re here to help.
Let’s unlock better analytics together.