Vitess Release Management

23 posts · Mar 2020 – Nov 2025

What this trail covers

Vitess 20 enhances query compatibility with improved DML support, Vindex hints, and advanced sharded update/delete operations. VReplication gains experimental multi-tenant import capabilities. Online DDL sees improvements in enum reordering, instant DDL prediction, range partitioning, and VDiffs on in-progress workflows, while dropping support for gh-ost. The Vitess Operator introduces experimental automated and scheduled backups.

All PlanetScale trails

Nov 2025Announcing Vitess 23This post announces Vitess 23.0.0, detailing its new defaults (MySQL 8.4), improved operational tooling (VTOrc, topology controls), and refined metrics (VTGate's TransactionsProcessed, VTOrc's SkippedRecoveries). It also covers deprecations of legacy metrics and API endpoints, updates to VTTablet and CLI/Docker, and specific upgrade instructions for operator users migrating from MySQL 8.0 to 8.4. Future plans include deeper MySQL 8.4 compatibility and expanded observability for VReplication, MoveTables, and Resharding.
Apr 2025Announcing Vitess 22This post announces Vitess 22.0.0 and Vitess Operator 2.15.0. It details query serving enhancements like query execution plan metrics, prepared statement caching, and GA for sharded views/atomic distributed transactions. Cluster management improvements include VTOrc enhancements for stalled disk recovery, errant GTID discovery, and semi-sync monitoring. The Kubernetes Operator now uses VTBackup pods for automated backups and supports Kubernetes 1.32. Performance gains are highlighted through gRPC codec upgrades, AST walk reduction, AST rewriting optimization, and merge-sort avoidance on single-shard queries.
Jan 2025Automating cherry-picks between OSS and private forksThis post details the development and implementation of a Vitess cherry-pick bot to automate the process of synchronizing private Vitess forks with open-source changes. It describes the evolution from a manual approach to using `git-replay` and finally to a stateful bot leveraging GitHub Actions and a PlanetScale database. The bot's core functionalities include identifying and executing cherry-picks from OSS `main` to the private `upstream` branch, and backporting changes to private `latest` branches based on labels. It also outlines the conflict resolution strategy, PR assignment, and the implementation of weekly integrity checks for reconciliation and alerting.
Oct 2024Announcing Vitess 21Vitess 21 introduces experimental support for atomic distributed transactions and recursive CTEs, enhancing query compatibility. VReplication gains reference table materialization and dynamic workflow configuration. Cluster management is improved with new VTOrc metrics for errant GTIDs. The throttler now supports multi-metric configurations for finer control over throttling behavior. Online DDL receives various improvements, including better instant DDL analysis and programmatic charset conversion. Backup and restore processes gain an experimental mysqlshell engine for logical backups and restores. VExplain is enhanced with `vexplain trace` for detailed query execution insights and `vexplain keys` for analyzing sharding key usage. The Vitess Kubernetes Operator adds horizontal autoscaling for VTGate pods and support for Kubernetes 1.31. VTAdmin introduces new pages for VReplication workflow creation/management and distributed transaction dashboards.
Sep 2024Instant deploy requestsIntroduces 'instant deployments' for PlanetScale deploy requests, which leverage MySQL's INSTANT DDL algorithm for schema changes that qualify. This feature significantly reduces deployment time for eligible operations like adding columns, changing default values, creating/dropping tables, and modifying views. It is an opt-in feature with caveats including non-revertibility and potential table locks.
Advertisement
Aug 2024Zero downtime migrations at petabyte scaleThis post details PlanetScale's zero-downtime database migration process for petabyte-scale datasets, built upon Vitess's VReplication and `MoveTables` workflow. It outlines a 7-step migration process that includes taking consistent snapshots, continuous replication of changes, data verification via VDiff, a pre-cutover traffic routing phase (Application->PlanetScale->OldDB), transparent cutover with query buffering and pausing, and optional reverse replication for rollback. The post emphasizes that this process avoids downtime and data loss, and can be reversed if necessary. It also highlights that horizontal sharding is often a part of these migrations when databases exceed 250GiB.
Jul 2024The State of Online Schema Migrations in MySQLThis post provides a comprehensive overview of online schema migrations in MySQL, contrasting native `INPLACE` and `INSTANT` DDL with external solutions. It details the caveats of `INPLACE` (resource greedy, replication lag, uninterruptible) and `INSTANT` (limited scope, metadata-only changes, risks with `DROP COLUMN`). It then elaborates on the common design and characteristics of external tools like Vitess, gh-ost, and pt-online-schema-change, emphasizing their ability to handle almost all schema change types, respect production workloads, and offer interruptibility, while noting potential limitations with foreign keys and partitioning. The post concludes that for the vast majority of changes, external tools are the preferred solution, and suggests using them consistently even for simpler changes supported by `INSTANT` DDL to avoid complexity.
Jun 2024Announcing Vitess 20Vitess 20 release notes detailing enhancements in query compatibility (Vindex hints, sharded updates/deletes, multi-table updates), experimental multi-tenant imports for VReplication, improved Online DDL support (enum reordering, instant DDL prediction, range partitioning, in-progress VDiffs), and the removal of gh-ost support. The Vitess Operator v2.13.0 adds experimental automated and scheduled backups.
May 2024Self-managed Vitess vs Managed Vitess with PlanetScaleThis post details the significant resources, time, and expertise required to self-manage Vitess, citing examples from Slack and Square to illustrate the complexities of implementation, migration, and ongoing maintenance. It highlights challenges such as MySQL compatibility, sharding strategy, server provisioning, operational processes (deployment, service discovery, backup/restore, topology management), application integration, double-writes, and handling production issues like deadlocks and scatter queries. In contrast, it positions PlanetScale as a managed solution that simplifies Vitess cluster creation and enhances the developer experience with features like database branching and deploy requests, thereby reducing the operational burden compared to self-hosting.
Apr 2024Summer 2023: Fuzzing Vitess at PlanetScaleThis post details the development of a random query fuzzer for Vitess's planner. The fuzzer randomly selects tables and columns to generate queries with clauses like SELECT, GROUP BY, WHERE, and ORDER BY. It also includes enhancements for random expression generation and query simplification, comparing results against MySQL to identify discrepancies. The work contributes to improving the robustness and correctness of Vitess releases by identifying and fixing bugs in the query planner.
Apr 2024How PlanetScale makes schema changesThis post details PlanetScale's internal workflow for making schema changes, which leverages their own online schema change tooling built on Vitess. It highlights the separation of schema changes from code deployments, the use of a GitHub Actions bot to automate branch creation and deploy requests, and the importance of the expand-migrate-contract pattern for backward compatibility. The post also emphasizes the queuing mechanism for concurrent schema changes and the safety checks performed by PlanetScale.
Dec 2023Introducing the schemadiff command line toolThis post introduces the `schemadiff` command-line tool, a wrapper around Vitess's `schemadiff` library. It enables engineers to validate, normalize, and diff database schemas from standard input, file system, or MySQL databases. Use cases include validating SQL syntax, prettifying schema definitions, and generating DDL for schema synchronization between branches. This tool can be integrated into CI/CD pipelines for pre-deployment validation.
Aug 2023Deploying multiple schema changes at onceThis post introduces the concept of near-atomic, multi-change deployments for database schema changes. It explains the technical challenges of applying multiple schema changes as a single unit, especially when dealing with long-running operations and inter-dependencies between changes. The post details how PlanetScale leverages Vitess's online schema change emulation and the `schemadiff` library to orchestrate concurrent execution of schema modifications, ensuring that all changes are staged and applied together, minimizing the window of schema inconsistency. It also discusses the benefits of this approach, such as reduced risk, improved control over deployment timing, and the ability to cancel deployments without impacting production.
Apr 2023Declarative schema migrationsThis post introduces the concept of declarative schema migrations, drawing parallels to Infrastructure as Code. It details how tools like the Atlas CLI can be used to define SQL schemas in configuration files, enabling automated application of schema changes. The post discusses the benefits (single source of truth, readability, automation) and drawbacks (potential for conflicts, acting as a crutch for DDL learning) of this strategy. It then outlines how this declarative approach can be integrated with PlanetScale's branching workflow and deploy requests, suggesting a dedicated 'db' repository for schema definition files and automation to apply these changes to development branches before production deployment.
Apr 2023Versioned schema migrationsThis post details how versioned schema migrations, a common pattern in application development (e.g., with Laravel's Artisan), can be applied to PlanetScale databases. It explains the mechanics of migration files, the `migrations` tracking table, and the `migrate` and `rollback` commands. It also outlines how to use these migrations with PlanetScale, differentiating between workflows with and without 'safe migrations' enabled, emphasizing the use of PlanetScale branches and deploy requests for production changes.
Mar 2023Announcing the PlanetScale GitHub ActionsThis post introduces the first wave of official PlanetScale GitHub Actions, including `setup-pscale-action`. These actions automate workflows for creating database branches, generating branch passwords, opening deploy requests, and commenting schema diffs on pull requests. Examples demonstrate how to integrate these actions into CI/CD pipelines for managing database schema changes alongside application code.
Sep 2022Gated Deployments: addressing the complexity of schema deployments at scaleIntroduces 'Gated Deployments' to address the complexity of multi-dimensional schema changes (multiple changes or multi-sharded databases). Gated Deployments allow for more atomic application of multiple schema changes by running lightweight tasks in parallel after bulk work is done sequentially, and by synchronizing changes across shards before a final, near-simultaneous application. It also introduces user-controlled deployment completion timing, allowing users to stage changes and manually apply them within a controlled window, extending the revert window and providing greater control over schema deployment timing.
Aug 2022Zero downtime Laravel migrationsThis post details how to achieve zero downtime Laravel migrations by leveraging PlanetScale's database branching and online schema change capabilities. It explains the dangers of direct DDL during deployment, the mechanics of non-blocking schema changes (shadow tables, atomic swaps), and provides detailed workflows for adding/dropping columns/tables and renaming columns/tables. It also covers the bonus feature of reverting schema changes within a 30-minute window by keeping the old table around and syncing data back, and clarifies that Laravel migrations should only be run on development environments connected to PlanetScale development branches, not production.
Jul 2022Getting started with the PlanetScale CLIThis post introduces the PlanetScale CLI, detailing its installation and usage for common database operations. It covers connecting to PlanetScale via `pscale login`, listing databases with `pscale database list`, creating databases with `pscale database create`, and entering a MySQL shell with `pscale shell`. It also demonstrates managing branches, including listing (`pscale branch list`), promoting (`pscale branch promote`), and creating (`pscale branch create`) branches. The post walks through creating a deploy request (`pscale deploy-request create`), listing deploy requests (`pscale deploy-request list`), and merging a deploy request (`pscale deploy-request deploy`), effectively showcasing the CLI's capabilities for managing schema changes and database workflows.
Jun 2022Announcing Vitess 14This post announces the general availability of Vitess 14, highlighting key improvements in usability and reliability. It details the GA status of Online DDL and the Gen4 planner becoming the default. Usability enhancements include command-line syntax deprecation warnings, the introduction of the new gRPC API for vtctld (VtctldServer) and its client (vtctldclient) with transition guidance, and the beta release of VTAdmin as a unified cluster management API and UI. Reliability improvements for VTOrc (still experimental) include better integration with Vtctld and federation handling, along with augmented Emergency Reparent Shard capabilities. Performance benchmarks show a ~10% improvement attributed to the removal of internal SAVEPOINT query execution.
Feb 2022Announcing Vitess 13This post announces Vitess 13, detailing its major themes: compatibility improvements (character sets, Gen4 planner), cluster management enhancements (VTOrc integration, errant GTID detection, robust emergency failovers), and better support for multi-column Vindexes with performance improvements for partial column routing. It also notes the introduction of versioned documentation on the Vitess website.
Sep 2020Orchestrator failure detection and recovery: New BeginningsThis post details the integration of Orchestrator into Vitess as a first-class citizen. It explains how the Vitess-aware Orchestrator leverages Vitess's topology server to understand cluster structure and expected states, enabling goal-driven failure detection and recovery. New scenarios addressed include replica without a primary, writable replicas, read-only primaries, and multi-primary setups, moving beyond the previous external integration's limitations.
Mar 2020PlanetScale migrates open source Vitess test suite from Python to GoMigrated the Vitess integration test suite from Python to Go, reducing development and debugging time, removing dependencies, and aligning with Python 2.7 end-of-life. Developed a Go test framework to programmatically interact with Vitess clusters, supporting parallel tests and non-conflicting working directories. Switched the CI pipeline from Travis CI to GitHub Actions to leverage larger compute instances and overcome concurrency limits, enabling more parallel test shards and faster feedback.