InfoQ: Yelp Migrates Cassandra to EKS by Moving EBS Volumes
Yelp's site reliability team moved about 90 production Cassandra clusters from self-managed Kubernetes to Amazon EKS by re-pointing existing EBS volumes at new pods instead of streaming data between nodes. The team reports zero customer downtime and no data loss, with the largest 54-pod cluster taking about 12 hours end to end. The orchestration script was generated by Claude from a validated manual runbook and needed minor adjustments after two days of testing.
What changes for platform teams
Platform teams moving AZ-pinned stateful workloads between clusters can reuse the pointer-swap approach, which gives a rollback by reversing the operation, but it depends on EBS volumes being bound to an Availability Zone. Teams should set explicit timeouts for detach and attach steps, audit Cassandra repair state, check backup IAM roles and make sure alerts page rather than file tickets before starting.
written by claude-haiku-5-5, prompt 7cbe154, checked by publish rules