Reliability fail: No automated zone failover for Coinbase’s global trading service

This article examines the major Coinbase trading outage on May 7, caused by a regional AWS availability zone failure. Despite being a $40B company, Coinbase's matching engine was pinned to a single AWS zone due to low-latency requirements, and they lacked automated failover to another zone. The outage lasted nearly 10 hours, with Coinbase directing customers to AWS status pages. The postmortem reveals deliberate architectural choices prioritizing latency over resilience, and the absence of automated cross-zone failover. The article provides a technical breakdown of availability zones vs. regions, and discusses reliability engineering lessons for high-frequency trading systems.