Confluent CCDAK Glossary: Kafka and Streaming Terms
April 13, 2026
Confluent CCDAK glossary of Kafka partitions, producers, consumers, schemas, and troubleshooting terms.
Use this glossary when Kafka producer, consumer, and schema terms start to blur together. Keep it beside the cheat sheet and resources instead of using it as a substitute for scenario practice.
| Term |
Short meaning |
| Offset |
Position marker for a record within one partition |
| Partition |
Ordered log shard inside a topic and the real boundary for ordering |
| Consumer group |
Coordinated set of consumers sharing work across partitions |
| Rebalance |
Redistribution of partition assignments across group members |
| Idempotent producer |
Producer mode that reduces duplicate writes during retries |
Acknowledgment (acks) |
Producer durability setting that controls write confirmation behavior |
read_committed |
Consumer isolation level that hides uncommitted transactional records |
| Transaction |
Kafka mechanism for grouping related writes atomically in supported workflows |
| Auto commit |
Consumer mode that records progress automatically with less semantic control |
| Retention |
Policy that controls how long Kafka keeps log data |
| Compaction |
Log cleanup mode that keeps the latest value per key |
| Schema Registry |
Central service for managing and validating message schemas |
| Serialization |
Converting data into bytes for transmission |
| Compatibility mode |
Rule controlling how schemas may evolve without breaking readers or writers |
| Key |
Record field used to determine partition choice and ordering behavior |
| Lag |
Difference between produced records and what a consumer group has processed |
| At-least-once |
Delivery pattern that may replay duplicates but avoids silent loss |
| Kafka Streams |
Stream-processing library and API layer for Kafka-based application logic |
| Kafka Connect |
Integration runtime for source and sink connectors |
Commonly confused pairs
| Pair |
Keep this distinction clear |
| offset vs partition |
record position versus the log shard itself |
| topic vs partition |
logical record collection versus the actual ordered shard |
| retry vs idempotence |
another send attempt versus duplicate-protection behavior |
| idempotence vs transaction |
safe retries for one producer versus atomic multi-write workflow |
transaction vs read_committed |
write-side atomicity mechanism versus read-side isolation choice |
| retention vs compaction |
time or size-based cleanup versus latest-key preservation |
| key vs value |
partitioning and ordering input versus payload content |
| consumer lag vs end-to-end latency |
unprocessed backlog versus total application delay |
| Kafka core vs Streams/Connect |
plain producer-consumer semantics versus higher-level platform capability boundaries |
Fast boundary reminders
| If the term really points to… |
Think of it as… |
| ordering and scaling |
partition, key, consumer group |
| producer semantics |
acks, retries, idempotence, batching |
| consumer progress |
offset, commit timing, rebalance, lag |
| stronger guarantees |
transaction, read_committed, compatibility mode |
| adjacent platform capability |
Schema Registry, Streams, Connect |
If the confusion is really about…
Revised on Sunday, May 10, 2026