SPQR 2.6.0 with Multicolumn Sharding

Hi! Our team has been hard at work recently, and we’re excited to share some major improvements in this release. Here are the highlights:

Enhanced [Re]balancing in SPQR

This release includes numerous improvements to key range movement and redistribution in SPQR. Key changes include:

• A safer way to redistribute key ranges using the new CHECK and APPLY options in the REDISTRIBUTE KEY RANGE command.
• If a move operation gets stuck or fails, you can now retry it using the new RETRY MOVE TASK GROUP command.
• Various balancing fixes to improve robustness and correctness.

Multi-column sharding

SPQR now supports multi-column sharding. To use this feature, define your distribution using the CREATE DISTRIBUTION and ALTER DISTRIBUTION ATTACH RELATION commands, passing a list of sharding column types and names.

AUTO INCREMENT for reference relations

For reference relations, it's now possible to auto-increment values (typically the id column) in a consistent way across all shards.

UUID column type

SPQR now supports the UUID data type in distributed relations.

Virtual processing

In some cases, SPQR can respond without querying any shards. This applies to simple queries like SELECT 1, SET ..., or SHOW statements. This feature is especially useful for health checks and monitoring.

👉 Check out the full release notes.

Huge thanks to @reshke, @EinKrebs, @diPhantxm, @Denchick, @SolinMaxim, and @miaarb for their contributions!

Stay tuned! 🚀