BlockBeats report: On May 9, Coinbase CEO Brian Armstrong stated that yesterday’s Coinbase outage was absolutely unacceptable. The root cause was an overheated rack in an AWS data center, which led to multiple cooling system failures. Our services are designed with redundancy against the failure of any single AWS Availability Zone (AZ), and most of our systems did operate as intended last night—but not all. Centralized trading platforms, optimized for low latency and customer co-location, failed to implement full redundancy. While it is possible to make trading platforms resilient to AZ failures, doing so would introduce latency and compromise customer co-location.
In light of this incident, Coinbase will reassess and weigh options to significantly reduce downtime when migrating availability zones. We thank the AWS and Coinbase teams for working through the night to mitigate the issue; a detailed technical post-mortem will be shared soon.