Confluent Apache Flink Certificate study plan

Map the exam blueprint to Flink's architecture by writing a one-page outline linking each exam objective to the Flink component that implements it (JobManager, TaskManager, Dispatcher, ResourceManager).

Set up a local Flink cluster in Docker and run the bundled word-count example end to end from the CLI.

Explain, in your own notes, the difference between bounded and unbounded streams, event time versus processing time, and where state lives during a job.

Install the Confluent CLI and create a trial Flink environment so you can submit a SQL statement to a cloud-hosted compute pool.

Build a Kafka topic with a schema in Schema Registry and register it as a Flink table using the Kafka connector with the appropriate format (Avro or JSON Schema).

Write and run at least five Flink SQL queries covering projection, filtering, aggregation, and a windowed GROUP BY over the same source table.

Compare the semantics of tumbling, hopping, and session windows by running one query of each and recording how many rows each emits.

Configure a watermark strategy on an event-time column and observe how late-arriving records are handled with and without an allowed lateness setting.

Implement a DataStream job in Java or Python that reads from Kafka, applies a keyed process function, and writes results back to a second Kafka topic.

Add keyed state to that job using ValueState and verify that state survives a checkpoint by restarting the job from a savepoint.

Configure checkpointing with an interval and exactly-once semantics, then inspect checkpoint statistics in the Flink UI.

Reproduce a failure by killing a TaskManager and confirm the job recovers without duplicate output.

Distinguish between the Table API and DataStream API by rewriting one of your week-3 transformations in the Table API and noting the tradeoffs.

Practice converting between DataStream and Table using the documented bridging methods, including handling of changelog streams.

Implement a temporal join between a fact table and a versioned lookup table and explain how the time attribute constrains the join.

Diagnose and fix at least two common errors, such as a missing watermark or a schema mismatch between the connector and the table definition.

Compare at-least-once, at-least-once with idempotent sinks, and exactly-once processing end to end, and identify which sink types support each guarantee.

Configure a Kafka sink with exactly-once delivery and verify transactional behavior by inspecting consumer offsets and output duplicates.

Measure the effect of parallelism and slot count on throughput by running the same job with two different configurations and recording records per second.

Tune a state backend setting, such as enabling incremental checkpoints, and observe the change in checkpoint size.

Deploy a Flink SQL statement as a persistent statement in Confluent Cloud and confirm it continues running after you close the CLI session.

Create a compute pool, scale it, and observe how the statement reacts to the change in resources.

Configure a connection to an external system using the appropriate connector and validate that data flows in both directions.

Practice reading job metrics and identifying backpressure using the Flink UI or the Confluent Cloud monitoring views.

Review Flink's savepoint mechanics and practice taking, restoring from, and disposing of a savepoint for a running job.

Explain how the Flink runtime handles a rescale operation and what constraints apply when changing parallelism from a savepoint.

Write a troubleshooting checklist that maps at least six symptoms, such as checkpoint timeouts or serialization errors, to likely causes and fixes.

Re-run two earlier exercises from memory without referring to notes, then compare your results against the originals.

Take two full timed practice exams or build a 60-question self-test from your notes, and score each attempt.

Review every missed question and write a corrected explanation for each one in your own words.

Re-read your week-1 blueprint and confirm you can explain every listed objective without notes.

Do a final pass over weak areas identified by your practice scores, spending the remaining time on the lowest-scoring topics.