Apache Fluss 1.0.02026-09-22
Apache Fluss 1.0.0 is the first major release since Fluss graduated to an Apache Top-Level Project. This release introduces the Fluss Gateway, brings native Rust, Python, and C++ clients into the main project, and expands storage, Lakehouse, Flink, Spark, and production operations capabilities.
📥 Downloads: https://fluss.apache.org/downloads/#apache-fluss-100
📖 Release blog: https://fluss.apache.org/blog/releases/1.0/
⚠️ Before upgrading from 0.9, read the 1.0 Upgrade Notes, especially the Lakehouse upgrade order, ACL changes, and AWS credentials provider migration.
---
🌐 Fluss Gateway
Features and Improvements
- Introduce Fluss Gateway as a stateless Rust REST service, with HTTP lifecycle management, health checks, and graceful shutdown. (#3963)
- Add cluster discovery and metadata APIs for databases, tables, and partitions. (#4078)
- Add REST APIs for database, table, and partition DDL operations. (#4107)
- Add schema-aware JSON decoding and batch append, upsert, partial update, and delete operations with per-row outcomes. (#4104, #4126)
- Add configuration validation and secret redaction. (#3970)
- Add Prometheus identity labels and improve REST access logs. (#4219, #4142, #4163)
- Provide Gateway binary distributions and container release support. (#4098, #4306)
Bug Fixes
- Reject duplicate MAP keys in JSON entry arrays. (#4205)
> Gateway is a preview in 1.0. It supports metadata, DDL, and writes; record reads and HTTP caller authentication are not included. See the Gateway guide for deployment requirements and API limitations.
---
🗄️ Server / Coordinator / Storage
Features and Improvements
- Add Coordinator high availability, leader election, and coordinator epoch fencing. (bbc445a, c49b848, #2781)
- Keep standby Coordinators synchronized with dynamic configuration changes. (#3645)
- Add a cluster health API for checking leaders and in-sync replicas during rolling upgrades. (#3400)
- Support multiple local data directories on TabletServers. (#3030)
- Add multiple remote data directory configuration and directory selection when creating tables and partitions. (#2757, #2763, #2769)
- Add disk-usage write protection with a configurable recovery threshold; prevent KV leader promotion while disk writes are blocked. (#3340, #3612, #3539)
- Add resource-based admission control for KV leader replicas. (#3614)
- Add cooperative KV write backpressure based on RocksDB Level-0 file pressure. (#3463)
- Share RocksDB block cache and write-buffer memory across KV tablets. (#2627, 059d7aa)
- Add row-level TTL for primary key tables through table.kv.ttl. (#3621)
- Support independent local log retention with table.log.local-ttl and dynamic updates to table.log.ttl. (#3778, #3233)
- Support rolling expired active log segments on low-traffic tables, with dynamic configuration updates. (#3831, #3940)
- Add V1 log batches with column statistics and server-side predicate pruning. (#2886, #2951)
- Add server-side KV scans for reading primary key tables in batches. (#2976, #3151)
- Support changing the default bucket count for future partitions while preserving existing partition layouts. (#3908, #4315)
- Support adding aggregation columns to existing tables. (#3459)
- Support standby replica election for primary key tables and table-level standby replica configuration; disable standby replicas by default for newly created primary key tables. (#2829, #3257, #4157)
- Add table-drop throttling to protect the Coordinator event queue. (#3174)
- Add per-tablet-server request retries and a dedicated executor for leader/ISR notifications. (#3391, #3647)
- Batch leader/ISR updates during controlled shutdown and require the new leader's acknowledgment for leader-only rebalancing. (#3654, #3556)
- Add timeouts for rebalance tasks and serialize leader-only rebalance operations. (#3429, #3071)
- Improve remote log recovery with asynchronous prefetch and remote CDC segment downloads. (#3132, #2855)
- Periodically checkpoint log recovery points. (#3547)
- Speed up shutdown by closing KV tablets concurrently and skipping RocksDB flushes during TabletServer shutdown. (#3651, #3807)
- Return ISR and bucket epoch information in bucket metadata. (#3785)
- Support dynamic updates to auto-partition enablement, precreation count, and retention count; add the yyyy-MM-dd partition time format. (#3453, #3434, #2964, #3558)
- Support altering database comments and custom properties. (#1172)
- Support dynamic updates to the minimum ISR configuration and KV snapshot interval. (#2837, #2820)
Bug Fixes
- Fix log recovery failures after unclean shutdown, orphan segments after full truncation, and writer recovery with sequence gaps. (#2942, #4013)
- Recover log segments contai
Apache Fluss 0.9.1-incubating2026-05-04
> DISCLAIMER: Apache Fluss (incubating) is an effort undergoing incubation at The Apache Software Foundation (ASF), sponsored by the Apache Incubator PMC.
>
> Incubation is required of all newly accepted projects until a further review indicates that the infrastructure, communications, and decision making process have stabilized in a manner consistent with other successful ASF projects.
>
> While incubation status is not necessarily a reflection of the completeness or stability of the code, it does indicate that the project has yet to be fully endorsed by the ASF.
Apache Fluss 0.9.1-incubating is a patch release on the 0.9 line containing bug fixes, Helm chart improvements, dependency upgrades, and documentation updates. It is fully compatible with 0.9.0 — no client or server changes are required to upgrade.
📥 Downloads: https://fluss.apache.org/downloads/#apache-fluss-091
---
🐛 Bug Fixes
Server / KV / Storage
- Fix Paimon IOManager leak in MergeTreeWriter causing /tmp disk exhaustion during tiering (#3193)
- Fix non-target columns not being nulled during partial update on first insert (#3157)
- Fix OutOfBoundsException in partial update / delete after ADD COLUMN (#3090)
- Fix ListDatabaseSummary always returning a table count of 0 or 1 (#2899)
- Still remove server tag if the tablet server no longer exists (#2884)
- Validate non-primary-key tables for KvSnapshotLease acquire / release APIs (#2736)
Client
- Fix early termination in SortMergeReader on changelog deletes (#3137)
- Catch Throwable to avoid silent failure on remote file download (#3136)
Flink Connector
- Fix $changelog and $binlog virtual tables failing with NPE on checkpoint completion (#2743)
- Fix release of KV snapshot lease for log tables (#2721)
- Fix acquire of KV snapshot lease for legacy clusters (#2720)
Filesystem
- Propagate S3 path-style-access config from server to client via delegation token (#3165)
---
✨ Improvements
- [flink] Push down lake filters for non-partitioned scans (#3181)
- [flink] Allow partial updates for the AGGREGATION merge engine (#2905)
- [server / netty] Introduce netty.server.max-request-size to bound request size (#3108)
- [filesystem / s3] Support the default AWS credential chain for S3 delegation (#3094)
- [filesystem / rustfs] Support AssumeRole STS for RustFS (#2989)
---
⛵ Helm Chart
- Bump chart version to 0.9.1-incubating (#3205)
- Add existingSecret support for SASL credentials (#3189)
- Support injecting environment variables and external secrets (#3184)
- Add pod annotations, labels, and PodDisruptionBudget support (#3185)
- Add scheduling primitives (nodeSelector / tolerations / affinity) (#3175)
- Backport general Helm chart improvements to release-0.9 (#3183)
- Reintroduce the Quickstart image (#3166)
---
📦 Dependency Upgrades
- Bump org.assertj:assertj-core 3.23.1 → 3.27.7 (#3119)
- Bump org.apache.logging.log4j:log4j-1.2-api 2.17.1 → 2.25.4 (#3118)
- Bump com.google.protobuf:protobuf-java 2.5.0 → 3.25.5 (#3112)
- Bump org.apache.kafka:kafka-clients 3.9.1 → 3.9.2 (#3111)
---
📚 Documentation
- Add "Insert If Not Exists" section to Lookup Joins documentation (#3037)
- Fix RustFS AssumeRole documentation (#3058)
---
🔗 Full Changelog
v0.9.0-incubating...v0.9.1-incubating