Flink Mastery (Senior → Principal)

Operate

• Start from correctness model: event time, ordering, lateness, replay, and side-effect semantics.
• Treat Flink as a stateful distributed system, not just a transformation engine.
• Prefer explicit state, windowing, and failure behavior over vague stream magic.
• Optimize for correctness first, then latency, then cost.

Default Standards

• State and checkpoint strategy must be explicit.
• Event-time assumptions should be validated with real lateness patterns.
• Backpressure and operator chaining require observability.
• External side effects must be designed for retries and replay.
• Multi-tenant stream platforms need strong workload isolation thinking.

References

• Stream architecture and operator design: references/stream-architecture-and-operator-design.md
• State, checkpoints, and savepoint safety: references/state-checkpoints-and-savepoint-safety.md
• Event time, watermarks, and lateness correctness: references/event-time-watermarks-and-lateness-correctness.md
• Backpressure, scaling, and performance tuning: references/backpressure-scaling-and-performance-tuning.md
• Delivery semantics and external side effects: references/delivery-semantics-and-external-side-effects.md
• Multi-tenant governance and stream platform safety: references/multi-tenant-governance-and-stream-platform-safety.md
• Reliability and operations: references/reliability-and-operations.md
• Incident runbooks: references/incident-runbooks.md