Message Storage Migration to Vitess

This thread tracks the evolution of Slack's message storage capabilities, focusing on scaling and performance improvements. Initial posts detailed the challenges of handling massive message volumes and the need for a more robust storage solution. This post details the technical work involved in migrating the core messages table to a Vitess cluster, a distributed MySQL sharding solution, to improve scalability for large customers. It covers the end-to-end process, from message sending to persistence, and the migration from an active-active MySQL configuration to Vitess, which now serves 99% of Slack's query load and handles 2.3 million QPS at peak with a median latency of 2ms.