Application Observability – CCDAK Study Guide
This domain accounts for 15% of the exam and covers monitoring Kafka client and Streams applications in production — the metrics, logs, and traces that tell you whether your producers and consumers are healthy. The single most important concept is consumer lag; know what it means, how to read its trend, and how to alert on it. The rest is the three pillars of observability applied to Kafka.
Consumer Lag — the Central Metric
Consumer lag = the difference between a partition's latest produced offset (log-end offset) and the consumer's committed offset. It measures how far behind a consumer is.
- Low and stable lag (small oscillations as batches process) = healthy; consumers keep up.
- Continuously growing lag = consumers are processing slower than producers are producing and are falling behind — scale consumers/partitions, optimize processing, or add capacity.
Lag is the primary health signal for a consumer group and the most common exam question in this domain.
Metrics via JMX
Kafka clients (producers, consumers, Streams) expose runtime metrics through JMX, which monitoring systems scrape (often via exporters). Key metrics:
| Metric | What it reveals |
|---|---|
| Consumer lag (per partition) | Whether consumers keep up |
| Producer request latency | How long produce requests take to be acknowledged |
| Producer record error rate | Records failing to be produced |
| Offset commit rate / latency | Whether consumers commit progress (not stuck reprocessing) |
| Throughput (records/bytes per sec) | Load and capacity |
Rising producer latency or error rate signals broker, network, or availability issues to investigate.
Unlock with a paid plan
4 more sections (about 258 words) continue below. Every paid plan unlocks all CCDAK study guides, the full question bank and mock exams, from $4.97/month.