Skip to main content

v1.0.0-rc.1

Β· 9 min read

Release date: February 02, 2026

πŸ‘ Highlights​

Region Repartition RC.1 introduces Region Repartition, allowing users to dynamically adjust partition rules and redistribute data at runtime, without rebuilding tables or performing manual migrations.

  • Split Partitions Split a large or hotspot partition into multiple smaller ones:
ALTER TABLE sensor_readings SPLIT PARTITION (
  device_id < 100
) INTO (
  device_id < 100 AND area < 'South',
  device_id < 100 AND area >= 'South'
);
  • Merge Partitions Merge multiple small partitions into a single one:
ALTER TABLE sensor_readings MERGE PARTITION (
  device_id < 100 AND area < 'South',
  device_id < 100 AND area >= 'South'
);
  • Execution Control with WITH Clause Users can control procedure behavior via WITH options:
ALTER TABLE sensor_readings SPLIT PARTITION (
  device_id < 100
) INTO (
  device_id < 100 AND area < 'South',
  device_id < 100 AND area >= 'South'
) WITH (
  TIMEOUT = '5m',
  WAIT = false
);
  • When WAIT = false, the statement returns immediately with a procedure_id. Progress can be checked via ADMIN procedure_state(procedure_id);
  • TIMEOUT sets an overall time limit for the operation.

Note: Region repartitioning is currently supported only when:

  • Running in distributed cluster mode
  • Using object storage and GC are enabled

Metric Engine Primary Key Filter Acceleration

RC.1 introduces a fast-path optimization for primary key filtering in the Metric Engine. Instead of decoding and materializing values row by row, the engine now performs direct comparisons on memcomparable-encoded primary key bytes, significantly reducing CPU overhead.

  • Dense codec: 20–90Γ— faster
  • Sparse codec: 3–11Γ— faster

This improvement greatly boosts query performance for high-cardinality metric workloads.

Breaking changes​

⚠️ Compatibility notice​

  • refactor: remove region_numbers from TableMeta and TableInfo by @WenyXu in #7519

In the new version, the TableMeta.region_numbers field has been removed. As older versions still depend on this field, downgrading from the new version to an older version may introduce compatibility issues. Specifically, an older version may fail or throw errors when reading data generated by the new version, since the region_numbers field is no longer present.

πŸš€ Features​

  • feat: impl vector index building by @killme2008 in #7468
  • feat: update ApplyStagingManifestRequest to fetch manifest from central region by @WenyXu in #7493
  • feat: impl json_get_int for new json type by @MichaelScofield in #7495
  • feat(meta-srv): fallback to non-TLS connection when etcd TLS prefer mode fail by @WenyXu in #7507
  • feat: implement RemapManifest and ApplyStagingManifest for repartition procedure by @WenyXu in #7509
  • feat: ingest jsonbench data through pipeline by @MichaelScofield in #7312
  • feat: repartition grammar candy by @waynexia in #7518
  • feat: adjust some args to gc worker by @discord9 in #7469
  • feat(repartition): implement region deallocation for repartition procedure by @WenyXu in #7522
  • feat: bump promql-parser to v0.7.1 by @waynexia in #7521
  • feat: tune query traces by @waynexia in #7524
  • feat: update special remote write label name by @sunng87 in #7527
  • feat(repartition): implement region allocation for repartition procedure by @WenyXu in #7534
  • feat(repartition): implement validation logic for repartition table by @WenyXu in #7538
  • feat: integrate repartition procedure into DdlManager by @WenyXu in #7548
  • feat: make grpc can handle metric engine request by @fengjiachun in #7508
  • feat: always canonicalize partition expr by @waynexia in #7553
  • feat: gc schd update repart mapping by @discord9 in #7517
  • feat: impl vector index scan in storage by @killme2008 in #7528
  • feat: add per-partition convert, result cache metrics by @evenyag in #7539
  • feat(cli): support RaftEngine store backend for metadata snapshot operations by @WenyXu in #7467
  • feat: enable pruning for manipulate plans by @waynexia in #7565
  • feat: add sync region instruction for repartition procedure by @WenyXu in #7562
  • feat: flush region before close when skip-wal is enabled by @v0y4g3r in #7549
  • feat: support for types of series identifier in series divide plan by @waynexia in #7583
  • feat: export import v2 rfc by @fengjiachun in #7499
  • feat: debug/mem/symbol by @discord9 in #7573
  • feat(meta): add repartition timeouts and metrics by @WenyXu in #7587
  • feat: exact partition filter by @discord9 in #7571
  • feat: add ddl timeout/wait options, repartition WITH parsing, meta-client startup refactor by @WenyXu in #7589
  • feat: gc worker on dropped region by @discord9 in #7537
  • feat: use tsid on promql planner by @waynexia in #7590
  • feat: implement view-based vector types by @waynexia in #7600
  • feat: ser/de for HistogramFold&UnionDistinctOn by @discord9 in #7601
  • feat: support TQL parsing in CREATE TRIGGER by @fengys1996 in #7599
  • feat: UDF json_get with user specified return type by @sunng87 in #7554
  • feat: ensure datanode plan's dist and ordering for tsid col by @waynexia in #7614
  • feat: add partial_drop to DropRequest by @WenyXu in #7597
  • feat: partition rule simplifier by @waynexia in #7622
  • feat: use arrow-pg for postgres data encoding by @sunng87 in #7591
  • feat: impl vector index query by @killme2008 in #7564
  • feat: BulkMemtable compact parts without encoding into Parquet by @evenyag in #7617
  • feat: flow last non null by @discord9 in #7646
  • feat: add recover_sync to ManifestCache::new by @evenyag in #7652

πŸ› Bug Fixes​

🚜 Refactor​

  • refactor: remove region_numbers from TableMeta and TableInfo by @WenyXu in #7519
  • refactor: refactor CreateTableProcedure to extract reusable components by @WenyXu in #7526
  • refactor: remove duplications from mito by @waynexia in #7632
  • refactor: SQL and gRPC server handlers by @waynexia in #7637

⚑ Performance​

🎨 Styling​

πŸ§ͺ Testing​

  • test(cli): fix snapshot test paths on Windows by @WenyXu in #7574
  • test: add integration tests for repartition by @WenyXu in #7560

βš™οΈ Miscellaneous Tasks​

All Contributors​

We would like to thank the following contributors from the GreptimeDB community:

@MichaelScofield, @Standing-Man, @WenyXu, @discord9, @evenyag, @fengjiachun, @fengys1996, @killme2008, @shuiyisong, @sunng87, @v0y4g3r, @waynexia