Vitess 12 Release: New Gen4 Planner, VTAdmin Enhancements, RBAC, and Benchmarking Improvements

Vitess 12 official version has been released.

Vitess is a database solution for deploying, scaling, and managing large MySQL instance clusters, combining many important MySQL features with NoSQL‑style scalability. Its architecture lets you run it efficiently in public or private clouds as if on physical machines, and it helps solve the following problems: Support sharding MySQL databases to scale them with minimal application changes. Migrate from physical machines to private or public clouds. Deploy and manage large numbers of MySQL instances.

In this version, Vitess makes major advances in several areas, including the Gen4 planner, VTAdmin, and other improvements.

Gen4 Planner

Gen4 is the latest version of the query planner, introduced as an experimental feature in Vitess 12. To use Gen4, the -planner_version flag of VTGate must be set to gen4 .

VTAdmin

The experimental multi‑cluster management API and Web UI introduced in Vitess 10.0, called VTAdmin, now includes improvements to the vreplication‑based Reshard workflow.

Vitess 12.0 introduces an experimental role‑based access control (RBAC) implementation, allowing Vitess operators to allow or deny API endpoints based on specific authorizations in their Vitess environment. This lays the groundwork for the upcoming deprecation of the vtctld2 UI. Note that VTAdmin does not provide any authentication implementation; users must supply their own authentication suitable for their deployment.

Deploying the vtadmin‑api and vtadmin‑web components is completely optional. VTAdmin depends on the new VtctldServer API, so the new grpc‑vtctld service must be running on vtctlds to use it.

Benchmarking

According to the official description, since the last release, arewefastyet has undergone subtle changes. The web server now uses a new benchmark queue that consumes fewer compute resources and avoids running the same benchmark twice. To increase confidence in the new Gen4 planner’s performance, the team added a feature that visualizes query plans and statistics (execution time, execution count) generated by macro‑benchmarks, providing more advantages when comparing V3 and Gen4 performance.

More Inclusive Naming

This version includes significant naming changes, such as removing the term “master” and replacing it with “primary” or “source”; these changes are now backward compatible. In the next version, deprecated commands will be removed, meaning scripts using deprecated commands should be updated to use the new ones.