// the find
GoogleCloudPlatform/flink-on-k8s-operator
[DEPRECATED] Kubernetes operator for managing the lifecycle of Apache Flink and Beam applications.
A Kubernetes operator for managing Apache Flink clusters and jobs via a FlinkCluster CRD. It handles the full lifecycle: job submission, savepoints, restarts from savepoint on failure, and GCS/GCP integration. Deprecated since September 2022 in favor of the official Apache Flink Kubernetes Operator.
The CRD design is well thought out — separating job clusters from session clusters at the spec level is the right call. Savepoint management (periodic, on-demand, cancel-with-savepoint) is genuinely useful and covered end-to-end. Volcano batch scheduler integration for co-scheduling JobManager and TaskManager pods is a real operational need that most alternatives ignore. The controller split into observer/reconciler/updater is clean Go operator structure.
It's deprecated — the README's first line tells you to go somewhere else, and the last commit was 2022. The API is v1beta1 with explicitly no backward compatibility guarantees, so anything built on it is a migration risk. No Helm chart for the operator itself in the initial release (only added late, and underdocumented). The Apache Flink Kubernetes Operator from the Flink project has since overtaken this in features and is actively maintained.