Zero downtime Upgrade: Yelp’s Cassandra 4.x Upgrade Story

Yelp's Database Reliability Engineering team upgraded over a thousand Cassandra nodes from version 3.11 to 4.1 with zero downtime. The post covers their upgrade strategy including benchmarking (4% p99 latency improvement, 11% throughput gain, up to 58% p99 reduction on key clusters), compatibility challenges with Stargate proxy and Cassandra Source Connector, and a three-stage automated upgrade process (pre-flight, flight, post-flight). Key lessons include a Stargate 2.x regression causing slower range queries (resolved by downgrading to 1.x), schema disagreement on CDC-enabled clusters, and the value of running version-specific components in parallel during the transition. The upgrade was performed in-place via rolling restart on Kubernetes, avoiding the cost and complexity of a separate DC approach.