Jean-François Gagné
I am Jean-François (J-F), a System Engineer and MySQL Expert with 23 years of experience, recognised as a MySQL Rockstar (2024) and MySQL Community Contributor Award recipient (2018). Over the last 12 years, I have solved reliability and scalability challenges at Booking.com, HubSpot, MessageBird, Aiven, and Proton. Before MySQL, I worked as a Linux SysAdmin, Telco Architect, and Programmer. At university, I studied Cache Consistency in Distributed Systems and Group Communication Protocols.
Proposals - Montreal 2027
The Dark Side of Auto-Increments (and SERIAL & Sequences)
MySQL Auto-increments feel like an obvious choice. But after 12 years of managing databases, I have seen the problems they can create at scale. From mistaken assumptions and minor coding mistakes to production incidents, we will examine how Auto-Incs create unexpected challenges. We will cover strategies to avoid these pitfalls and safer alternatives to Auto-Incs. Some of the lessons may apply to sequences in PostgreSQL and MariaDB.
All I Wanted to Know About Sharding
Covers sharding functions beyond simple modulo and hashing, vertical (split) and horizontal sharding, rebalancing and resharding, and post-sharding reliability considerations. Lessons are drawn from 12 years of working with MySQL at scale and are broadly applicable to other databases. Will discuss considerations for Vitess and TiDB, many of which are likely also applicable to PostgreSQL Neki and other distributed solutions.
Autopsy of a MySQL Automation Disaster
A MySQL database failed, and our automation correctly kicked in. But a rare failure, human error, edge-case recovery, and subtle tooling flaw later turned this routine failover into a split-brain. This talk breaks down the exact timeline of the incident, why it happened, and lessons for prevention and recovery. We will explore ideas for building safer automation, sharpening incident response, and designing more resilient database systems.
MySQL Scalability and Reliability for Replicated Environment
Shifting from a single MySQL server to a replicated environment sounds simple, but scaling out brings challenges. We will cover managing a replicated environment, including adapting applications for replication-friendly operations, coping with replication lag, achieving read-your-own-writes on replicas, and enabling quick primary failover. We will also expose the pitfalls of UNIQUE and FOREIGN KEYs, as well as those of Auto-Increment.
Autopsy of a Cascading Outage caused by a MySQL Crashing Bug
Once upon a time, a query triggered a crashing bug. After automated failure handling, the application re-sent the query, and MySQL crashed again. This triggered a cascading failure, which resulted in datastore unavailability. Unfortunately, mistakes made during incident response led to avoidable data loss. This talk covers what happened, how we handled the incident, what we learned, and the infrastructure changes we made to prevent recurrence.
