Opening Scenario
Your team is preparing to move from a single Flink dev cluster to production. The platform team asks how you want it deployed: "session mode on a shared cluster," "application mode per job," "per-job mode on YARN," or "reactive mode for autoscaling." The team has 8 jobs in active development, two of which will go to production this quarter. Each has different reliability requirements: one is a low-traffic side-pipeline that fits in 4 task slots; another is a 200-slot fraud-scoring job that has its own SLA. You need to understand the deployment modes Flink offers, the operational implications of each, and the right default for production streaming.
Question Progression
1. What is a "deployment mode" in Flink, and why does it matter?
Question. Deployment mode is the configuration of how the JobManager and TaskManagers are launched, how the user job is submitted, and how they relate. Why is this a first-class decision rather than an implementation detail?
In the full tutorial
- When to Reach for This