CEP-21: Practical Guide to TCM - New Paxos Based Metadata Manager

All we need to say is that Patrick McFadin is speaking. :) Join us as we explore the new way to run global scale clusters.

What You’ll Learn:
Cassandra has managed cluster metadata through gossip since the beginning. With Cassandra 6, that changes. Transactional Cluster Metadata (CEP-21) replaces gossip-based metadata propagation with a Paxos-backed distributed log, eliminating an entire class of operational pain points. Split-brain risk during topology changes, ring-settle guesswork, and inconsistent schema propagation.

Why You Shouldn’t Miss It:

This talk is a practical operator's guide to TCM. We'll cover what actually changes under the hood (and what doesn't), walk through the three-phase upgrade path and its point of no return, and show how day-to-day operations like bootstrap, decommission, and schema changes become faster and safer. We'll also cover CMS sizing and monitoring for production, failure playbooks for when things go wrong, and a concrete testing checklist you can use before enabling TCM in your environment.

If you run Cassandra clusters and want to understand how to plan for, upgrade to, and operate with TCM, this talk is for you.

About the Speaker:
With over 15 years of experience, including roles as an Oracle DBA/Developer and Chief Architect at Hobsons, Patrick McFadin has dedicated his career to solving complex distributed systems challenges. As the Chief Evangelist for Apache Cassandra at DataStax, he guides organizations in building some of the largest and most exciting production deployments. He is a co-author of "Managing Cloud Native Data on Kubernetes" for O'Reilly and travels globally to educate developers on distributed databases, streaming systems, and the future of data infrastructure. Outside of his work with Cassandra, Pulsar, Spark, and Flink, he can be found exploring retro computing or pursuing astrophotography.