Server-initiated instant failover in MariaDB
Abstract
Zero downtime upgrades in MariaDB traditionally require complex replication configurations, custom client orchestration, database proxies, or DNS/IP address manipulations. But what if achieving an instant failover to another MariaDB server could be as straightforward as a web server’s HTTP 302 redirect?
This talk presents a working prototype for server-initiated instant failover in MariaDB. By eliminating the need for elaborate replication setups and client configuration updates, this solution enables database administrators to seamlessly transition workloads between servers — whether during planned upgrades or emergency scenarios.