Confluent Certified Developer for Apache Kafka (CCDAK) Practice Exam

What the Confluent Certified Developer for Apache Kafka (CCDAK) exam covers

• Apache Kafka Application Development210 questions
• Apache Kafka Fundamentals173 questions
• Kafka Connect113 questions
• Application Observability98 questions
• Apache Kafka Streams90 questions
• Application Testing67 questions

Free Confluent Certified Developer for Apache Kafka (CCDAK) sample questions

A sample of 10 questions with answers and explanations. Sign up free to practice all 751.

1. Question 1Apache Kafka Application Development

   Which describes the role of a Kafka producer?

   • AIt publishes records to a topic, choosing or deriving the target partitionCorrect
   • BIt reads records from a topic and tracks its position with offsets
   • CIt stores the partitions of a topic on disk and serves fetch requests
   • DIt coordinates which consumer in a group owns each partition
   ✓ Correct answer: A

   Partition selection happens on the producer side rather than on the broker, which is why the key and the partitioner configuration decide where a record lands and therefore what ordering guarantee it receives. Reading, storing and group coordination are the consumer, broker and coordinator roles.

   Why the other options are wrong

   • BReading records and tracking position with offsets is the consumer's role.
   • CStoring partitions on disk and serving fetches is the broker's role.
   • DCoordinating partition ownership within a group is the group coordinator's role.
2. Question 2Apache Kafka Application DevelopmentSelect all that apply

   Which TWO should accompany a dead letter topic in production? (Choose TWO)

   • AMonitoring of the rate at which records arrive on itCorrect
   • BEnough context on each record to diagnose why it failedCorrect
   • CA retention period shorter than the source topic's
   • DThe same partition count as every other topic in the cluster
   ✓ Correct answer: A, B

   A dead letter topic nobody watches is a silent data loss mechanism, and a record with no error context tells an investigator nothing about what went wrong. Both are cheap to add. A shorter retention discards evidence sooner, and matching partition counts across the cluster is not a requirement.

   Why the other options are wrong

   • CA shorter retention than the source discards the evidence sooner.
   • DThere is no requirement to match the partition count of other topics.
3. Question 3Apache Kafka Application Development

   A producer's throughput is far below expectation although the cluster is lightly loaded. Which setting is the first to examine?

   • Alinger.ms, since a value of zero sends many tiny batchesCorrect
   • Bauto.offset.reset, since it governs where reading begins
   • Csession.timeout.ms, since it governs member liveness
   • Dmax.poll.records, since it governs batch size on read
   ✓ Correct answer: A

   A linger.ms of zero means the producer sends as soon as it can, so each request carries a handful of records and the per request overhead dominates, which is exactly the shape of a producer that is slow on an idle cluster. The other three are consumer settings or govern liveness rather than batching.

   Why the other options are wrong

   • BThe auto.offset.reset setting governs where a consumer starts reading.
   • CThe session.timeout.ms setting governs consumer liveness detection.
   • DThe max.poll.records setting governs the consumer's read batch.
4. Question 4Apache Kafka Fundamentals

   Which describes the active segment of a partition?

   • AThe segment currently being appended to, which is never deleted by retentionCorrect
   • BThe segment holding the records consumers are currently reading
   • CThe segment stored on the leader rather than on a follower
   • DThe segment that has been compacted most recently
   ✓ Correct answer: A

   Because retention deletes whole closed segments, the one still being written is always retained, which is why a low volume topic can hold records far older than its retention period suggests. Reader position, replica location and compaction recency are three other properties.

   Why the other options are wrong

   • BWhich segment consumers are reading depends on their position.
   • CBoth leaders and followers have an active segment.
   • DRecency of compaction does not define the active segment.
5. Question 5Apache Kafka Fundamentals

   Which happens when a new broker is added to an existing cluster?

   • AIt carries no partitions until a reassignment moves replicas onto itCorrect
   • BIt automatically takes an equal share of every existing partition
   • CIt immediately becomes the controller for the whole cluster
   • DIt rejects client connections until every topic is recreated
   ✓ Correct answer: A

   Existing partitions are not redistributed on their own, which surprises teams who add a broker expecting instant relief and see no change in load until the reassignment is planned and run. Automatic redistribution, automatic controller promotion and refusing connections are three things that do not happen.

   Why the other options are wrong

   • BThe new broker does not automatically take a share of existing partitions.
   • CBecoming the controller for the cluster is decided by election.
   • DClient connections are accepted rather than rejected pending recreation.
6. Question 6Apache Kafka Fundamentals

   Which happens when an infrequently running job finds its committed offsets expired?

   • AThe reset policy applies, so it replays from the start or skips to the endCorrect
   • BThe job is refused permission to join its consumer group again
   • CThe broker restores the offsets from the previous retention window
   • DThe job resumes from the offset it held before the expiry occurred
   ✓ Correct answer: A

   Neither outcome is usually what the job wanted, which is why a monthly batch consumer is either kept alive between runs or given an explicit starting position rather than relying on what was committed. Refusal, restoration and silent resumption are three things that do not happen.

   Why the other options are wrong

   • BThe job is not refused permission to rejoin its consumer group.
   • CThe broker does not restore offsets from a previous window.
   • DThe job cannot resume from an offset that has already expired.
7. Question 7Kafka Connect

   Which describes a router transformation on a sink connector?

   • AIt changes the destination topic name a record is treated as belonging toCorrect
   • BIt changes which broker the connector sends its fetch requests to
   • CIt changes the partition a record is assigned within its topic
   • DIt changes the consumer group the connector uses to read records
   ✓ Correct answer: A

   It is how records from many topics are collapsed into one destination table, or how a suffix is stripped before the name becomes a table name, without any change to the connector implementation. Broker selection, partition assignment and group naming are three other concerns.

   Why the other options are wrong

   • BWhich broker serves a fetch is decided by metadata rather than a transformation.
   • CThe partition a record occupies was fixed when the record was written.
   • DThe consumer group the connector uses is set in its configuration.
8. Question 8Application Observability

   Which consumer measurement shows time being lost inside the application rather than in Kafka?

   • AThe time spent between successive poll calls by the applicationCorrect
   • BThe time the broker takes to serve each fetch request
   • CThe number of records returned by the most recent poll
   • DThe number of members currently in the consumer group
   ✓ Correct answer: A

   A long gap between polls means the application is spending its time processing rather than waiting on Kafka, which is exactly the case where adding brokers or partitions changes nothing at all. The time the broker takes to serve a fetch, the number of records returned by a poll and the number of members in the group describe the cluster and the group rather than the application's own processing.

   Why the other options are wrong

   • BThe time the broker takes to serve a fetch measures the cluster instead.
   • CThe number of records returned by a poll describes volume, not time lost.
   • DThe number of members in the consumer group describes the group's size.
9. Question 9Apache Kafka Streams

   Which describes the relationship between tasks and partitions in a stream application?

   • AOne task is created per input partition, and tasks are spread over the instancesCorrect
   • BOne task is created per instance, regardless of the partitions involved
   • COne task is created per topic, whatever number of partitions it has
   • DOne task is created per processing step defined in the topology
   ✓ Correct answer: A

   It means the partition count is the ceiling on parallelism just as it is for an ordinary consumer group, and that adding instances beyond that number leaves them idle. A task per instance, a task per topic and a task per processing step are three incorrect models of how the work divides.

   Why the other options are wrong

   • BA task per instance would not divide the partitions between them.
   • CA task per topic would ignore how many partitions the topic has.
   • DA task per processing step is not how the work is divided.
10. Question 10Apache Kafka StreamsSelect all that apply

    Which TWO reduce the risk when changing a stream application's topology? (Choose TWO)

    • ANaming the operators and stores explicitly so names stay stableCorrect
    • BTesting the change against a copy of production data firstCorrect
    • CIncreasing the replication factor of the source topics
    • DReducing the number of instances before the deployment
    ✓ Correct answer: A, B

    Explicit names stop an unrelated edit from renaming internal topics and discarding state, and a rehearsal shows what the restore actually costs before it happens in production. Raising the replication factor of the source topics and reducing the instance count address neither the naming nor the rehearsal.

    Why the other options are wrong

    • CIncreasing the replication factor of the source topics does not help here.
    • DReducing the number of instances before deploying does not reduce this risk.

Who this Confluent Certified Developer for Apache Kafka (CCDAK) practice exam is for

This practice set is for anyone preparing for the Confluent Certified Developer for Apache Kafka (CCDAK) exam at the intermediate level - from first-time candidates building a foundation to experienced Data practitioners doing a final review before test day. If you learn best by working through realistic questions and reading why each answer is right or wrong, it is built for you.

How to use this Confluent Certified Developer for Apache Kafka (CCDAK) practice exam

1. Start with the free sample questions above to gauge your current baseline.
2. Read the full explanation on every question, including why each wrong option is wrong.
3. Track your weak domains and focus your study where you are losing the most marks.
4. Once you are scoring consistently well, take a timed, full-length mock exam.
5. Use your readiness score to decide when you are ready to book the real Confluent Certified Developer for Apache Kafka (CCDAK) exam.

Related Data resources

• Confluent Certified Developer for Apache Kafka (CCDAK) study guideKey concepts
• Data practice examsAll Data
• Certification pathWhere this fits
• Certification exam guides & tipsBlog
• Plans & pricingFree & paid
• Databricks Certified Data Analyst Associate practice examRelated
• CompTIA Data+ (DA0-002) practice examRelated
• MongoDB Associate Developer practice examRelated

Confluent Certified Developer for Apache Kafka (CCDAK) practice exam FAQ