MySQL Legend 2024. Q&A with Shlomi Noach

Q1. Congratulations for your title of MySQL Legend 2024. What is your role now that you are a Legend and how do you support the MySQL community?

I have been involved with the MySQL community for over 15 years. I’ve authored some widely used tools, such as orchestrator and gh-ost, written technical blog posts on all things database management and design, actively participated in conferences, and am a frequent speaker. I’m a strong believer in knowledge sharing and have long been an advocate for open source. For the past five years, I’ve been working at PlanetScale as a Vitess maintainer, further contributing to the MySQL ecosystem.

Q2. You are known for creating and maintaining Orchestrator. What is it, and what is it useful for?

Orchestrator is a MySQL high availability and topology management tool. It excels at analyzing replication topologies — who is replicating from whom, with what configuration, and in what state. It can:

  • Visualize your replication topologies and help you assess serving capacity across data centers or availability zones.
  • Refactor topologies — for example, to reduce cross-site traffic or offload replication from a server undergoing maintenance.
  • Detect errors: broken replication, unreachable nodes, configuration mismatches, and more. Notably, orchestrator uses a holistic approach to analyzing primary outages that is extremely reliable and noise-free.
  • Perform failovers and heal topologies. When a primary or other key server fails, orchestrator can rewire the topology and coordinate with external tools via hooks to enable full recovery — for example, informing Consul or etcd of a new primary.

Orchestrator itself operates in a highly available setup and can form a cross-availability-zone raft cluster. This ensures there’s always a node available for failovers and gives the cluster better insight into the state of the network.

Q3. Who is currently using Orchestrator and for what kind of applications?

Many companies use orchestrator at large scale. GitHub and Booking.com are two notable adopters, both of which contributed to its development. Numerous other well-known companies — including a major cloud provider — also use orchestrator to manage MySQL high availability, failovers, and replication topology.

Q4. You also worked on gh-ost, an online schema migration tool for MySQL. What was your contribution?

I co-designed, created, and authored gh-ost, and have maintained it as an open-source project, incorporating contributions from the community.

Q5. What are the benefits of such an online schema migration tool?

Most schema changes in MySQL are blocking (though MySQL’s INSTANT DDL and other improvements have helped). The common tool used was pt-online-schema-change, a trigger-based migration tool modeled after my earlier work, oak-online-alter-table.

At GitHub, schema changes were one of the most disruptive operations in production. The trigger-based approach added significant load, especially during the final cutover step — sometimes stalling tables or the entire database and causing repeated outages.

We envisioned a better solution: a trigger-less tool based on MySQL replication (CDC-style). We focused on operability — the ability to pause/resume at will, automatic throttling based on replication lag or other metrics, and a detached workload from production traffic. We succeeded in building that into gh-ost. After extensive testing, we adopted it in production and never looked back. Today, gh-ost is a widely used and capable schema change tool.

Q6. You are now the maintainer for the open-source Vitess. What is it?

Vitess is a CNCF-hosted, open-source, cloud-native framework for scaling MySQL. It’s best known for its advanced sharding capabilities. I’ve been maintaining Vitess for the past five years as part of the PlanetScale team.

Vitess enables horizontal sharding of databases to support massive data and traffic scale. Originally developed at YouTube to manage video metadata, it is now used by companies like Slack, where every message is served from and stored in Vitess.

Vitess presents itself as a single MySQL instance, while under the hood it rewrites and routes queries to the appropriate shard(s). It features a sophisticated query engine that supports cross-shard queries and transactions. It also enables live resharding, materialization, online schema changes, backups and restores, high availability, query caching, and more.

Q7. What is your vision for Vitess?

Over the years, I’ve worked at several companies managing their database infrastructure. Again and again, we had to solve the same problems. Across the MySQL community, many engineers invest significant time building and stabilizing infrastructure — even though that’s not their core business.

My vision is for Vitess to be the one infrastructure solution: everything built-in — high availability, backups, schema changes, resharding, and more. A single, comprehensive framework for managing production-grade MySQL at scale.

Q8. What do you see as future directions for MySQL?

While many database vendors compete on features, MySQL’s strength has always been operability — easy upgrades, flexible replication, backup and restore options, etc. These traits have enabled a rich ecosystem of community tools.

I believe further investment in operability will continue to benefit MySQL — both directly and by empowering the community to build on top of it.

Q9. Anything else you wish to add?

I hope to see continued and growing cooperation between MySQL and the open-source community.

 ……………………………………………………….

Shlomi Noach, Engineer and database geek at PlanetScale

Maintainer for open source Vitess. Previously at GitHub. Interested in database infrastructure solutions such as high availability, reliability, schema management, enablement, automation and testing. Shlomi is the author of orchestrator, gh-ost and other open source tools. He is the recipient of MySQL Community Member of the Year, Oracle ACE (Alumni) & Oracle Technologist of the Year awards.

Useful links:

Sponsored by MySQL/Oracle.

You may also like...