Rearchitecting Compute at Coinbase - Migrating To Karpenter for Fast, Reliable Scaling

Learn how Coinbase rearchitected their compute infrastructure by migrating from cluster-autoscaler and EKS managed node groups to Karpenter to handle thousands of workloads ranging from high-throughput APIs to long-lived blockchain nodes. Discover the specific pain points that drove this migration, including scaling latency issues, bin packing challenges, and the need to satisfy unique topology and hardware requirements in the volatile cryptocurrency environment. Explore the implementation of a scalable and tag-driven EC2NodeClass abstraction, understand how Karpenter was tuned for burst elasticity, and examine the observability and guardrails built to make scaling safer and more predictable. Gain insights from real-world lessons learned including challenges with overly aggressive node consolidation, launch template drift, VPC-CNI coordination issues, and how the platform was adapted to support rapid, responsive scaling under the intense pressure of crypto-scale traffic patterns where reliability and cost optimization are critical.