Recommended production topology:
- Flink cluster (Kubernetes via Helm)
- Flink Job Gateway as a Deployment/Service
- Kafka as source/sink (managed or self-hosted)
Key notes:
- Build and publish the Gateway; it prebuilds and bundles
flink-ir-runner.jarautomatically. - The Gateway uploads the runner jar to Flink on submission; no separate jar mounting step is required.
- Configure Gateway with
FLINK_CLUSTER_HOSTandFLINK_CLUSTER_PORT. - Use
parallelisminJobMetadatato scale jobs appropriately.
Security & reliability:
- Lock down Gateway with auth (API keys, mTLS) in front of cluster network.
- Enable Flink checkpoints and state backends suitable for durability.
- Add retries and idempotency around submissions (consider Temporal orchestration).