Lyft Moves Streaming Fleet to Apache Flink Kubernetes Operator

Lyft abandoned its custom Kubernetes operator for Apache Flink and migrated hundreds of production jobs to the open-source Apache Flink Kubernetes Operator. This enabled last-state upgrades, in-place autoscaling, and resource autotuning, saving several million dollars annually on excess capacity.

• Migration affected hundreds of production Apache Flink jobs
• Upgrades now use last-state without state loss
• Autotuning returned excess resources, saving several million dollars per year
• Lyft upgraded to Flink 1.19 for in-place scaling