Apply messaging, caching, retries, timeouts, circuit breakers, and idempotency.
Distributed systems · reliability
Use the strong answer as your reference, explain it aloud in your own words, then choose how soon you want to review it again.
Interview prompt
Strong answer
Kafka is a durable, partitioned log: ordering is guaranteed only within a partition, consumers in a group share partitions, and delivery is at-least-once unless you engineer otherwise, so consumers must be idempotent. Redis is an in-memory data store used for caching, rate limiting, distributed locks with leases, and short-lived state, and it should be treated as losable unless persistence is configured deliberately. Resilience patterns (timeouts, retries with backoff, circuit breakers, bulkheads, and dead-letter topics) keep partial failures from cascading.
Key points
Interviewer looks for
Explain the delivery guarantee you actually get, how duplicates and reordering are handled, and how retries, timeouts, and circuit breakers interact without amplifying load.
Java example
1@Component
2class PaymentEventsListener {
3 private final ProcessedEvents processed;
4 private final LedgerService ledger;
5
6 PaymentEventsListener(ProcessedEvents processed, LedgerService ledger) {
7 this.processed = processed;
8 this.ledger = ledger;
9 }
10
11 @KafkaListener(topics = "payments.captured", groupId = "ledger")
12 @Transactional
13 void on(PaymentCaptured event) {
14 if (!processed.markIfNew(event.eventId())) return; // idempotent consumer
15 ledger.record(event.orderId(), event.amount());
16 }
17}
18
19@Service
20class ShippingQuotes {
21 @CircuitBreaker(name = "carrier", fallbackMethod = "flatRate")
22 @Retry(name = "carrier")
23 @TimeLimiter(name = "carrier")
24 CompletableFuture<Quote> quote(Parcel parcel) {
25 return CompletableFuture.supplyAsync(() -> carrierApi.quote(parcel));
26 }
27
28 CompletableFuture<Quote> flatRate(Parcel parcel, Throwable cause) {
29 return CompletableFuture.completedFuture(Quote.flat(parcel));
30 }
31}Topic interview questions
These questions are tied to this lesson. Read each question aloud, then compare your explanation with the strong answer.
Strong answer
Only within a partition. Messages with the same key go to the same partition and are consumed in order by one consumer in the group. There is no global ordering across partitions, so the key must group the events that need to stay ordered, such as all events for one order.
Strong answer
A message can be redelivered after a crash or rebalance if its offset was not committed, so the same event may be processed twice. Consumers must be idempotent, for example by recording processed event ids in the same transaction as the side effect or using upserts.
Strong answer
Classify the failure, retry transient errors a bounded number of times with backoff, then publish the message with error metadata to a dead-letter topic and commit the offset so the partition is not blocked. Alert on DLT volume and provide a replay path.
Strong answer
A circuit breaker tracks failure rate or slow calls; when a threshold is exceeded it opens and fails fast for a period, then lets a few trial calls through in half-open state. Retries repeat a call hoping a transient error clears. Unbounded retries against a failing service amplify load, which the breaker prevents.
Strong answer
Acquire with SET key token NX PX ttl so the lock has an owner token and an expiry, release only if the token matches (atomically via a script), and keep the critical section shorter than the lease. For correctness-critical work, add a fencing token checked by the resource, because a paused client can outlive its lease.
How well do you know this?
Your review choice is saved locally and synced to your learner profile when signed in.