Yelp migrates Cassandra using zero-copy EBS disks

Yelp engineers describe how they successfully migrated their massive Apache Cassandra fleet of about 1,000 pods and a petabyte of data across Kubernetes clusters without downtime by moving Amazon EBS storage disks instead of streaming data over the network. By hardcoding EKS Persistent Volumes to point to existing EBS volume IDs and orchestrating automated pod handoffs, the team avoided weeks of network streaming and doubled compute costs. Open article →