GreptimeDB v0.6 Released - Support Migrating Table's Regions between Datanodes

As 2024 dawned, the Greptime team, invigorated by the New Year's fresh momentum, continued their efforts towards innovative version iterations. Just three weeks following our previous update, we are thrilled to announce a new version to our open-source time-series database: GreptimeDB v0.6.

This update marks a substantial leap from GreptimeDB v0.5, incorporating several major improvements.

GreptimeDB v0.6 Updates ​

Key Features ​

Region Migration

In version 0.5, we enabled support for Kafka WAL, making it possible to synchronize and migrate Region data across multiple Datanodes. In version 0.6, we initially implemented the Region Migration feature, providing users with the ability to migrate table Regions between Datanodes while ensuring data integrity. This lays a solid foundation for dynamically adjusting cluster load. For example, as query performance requirements increase, users can easily migrate table Regions to Datanodes with lower loads or larger specifications through Region Migration, achieving better query performance.

In the future, we plan to introduce dynamic Region distribution. This feature is designed to intelligently redistribute data Regions, leveraging real-time monitoring of workload conditions and business requirements while ensuring uninterrupted service. This strategic enhancement aims to optimize resource utilization. By doing so, it not only promotes more efficient and smarter data management but also ensures robust and adaptive support for the ever-evolving demands of the business environment.

Additional Updates ​

• Added a configuration item that allows specifying the default time zone for queries
  • By adding the --store-key-prefix configuration option, administrators can specify the Key prefix used by metasrv to avoid key name conflicts.
• Implemented the OR logical operator in PromQL
  • Added a special UNION operator (OR in PromQL) specifically for certain PromQL query scenarios. This operator takes two input nodes. All columns from the left child node will be output, and columns specified in compare_keys are used to check for duplicates. In case of duplicates, if both are from the right node, only the first row is retained; if from the left node, the row from the right node is discarded. The output includes columns from both left and right nodes, and the order of rows is not fixed.

Future Plans ​

Our next milestone, v0.7, promises to be even more exciting.

We plan to introduce a brand new indexing module, with its first implementation being an inverted index. This module aims to significantly boost performance when filtering and querying a small subset of time-series from vast datasets, a key focus for our Metric Engine in observable scenarios. Our team is currently rigorously testing the integration of these features to ensure optimal performance and stability. Stay tuned for the much-anticipated release of GreptimeDB v0.7!