This lesson matters because many Kafka application failures look similar at first. The exam wants you to classify the symptom correctly before you reach for a fix.
Symptom chooser
Symptom
Strongest first focus
lag rising
consumer throughput, partition count, downstream bottleneck
duplicate processing
retry path, commit timing, handler idempotence
out-of-order events
partition path, keys, in-flight retry behavior
producer timeout
broker path, acknowledgement expectations, or load pressure
What the exam is really testing
If the scenario shows…
Strong reading
duplicates after retry or restart
producer or commit behavior is under test
order problems for one entity
partition path and key behavior matter
backlog but stable group
throughput bottleneck may matter more than rebalance
timeout under strict guarantees
acknowledgement and load trade-offs may matter
Common traps
Trap
Better rule
blaming every duplicate on Kafka alone
application handling and commit timing often matter too
assuming order problems must be consumer bugs
producer partitioning or retry path may be the real cause
treating lag as proof that more consumers alone will solve it
downstream work and partition count still bound throughput
Decision order that usually wins
Classify the symptom first: lag, duplicates, ordering, and timeouts usually point to different primary lanes.
If one entity is out of order, inspect keys, partition routing, and retry path before consumer internals.
If duplicates appear after failure or restart, inspect commit timing and handler idempotence alongside producer retry behavior.
If timeouts appear under stricter guarantees, think acknowledgement expectations and load pressure before cosmetic tuning.
Quiz
This quiz requires JavaScript to run. The questions are shown below in plain text.
Loading quiz…
Revised on Sunday, May 10, 2026