Streaming & Resilience Checkpoint

Work fresh traces, diagnose invalid implementations, and identify model boundaries.

Get every question correct to pass this checkpoint. You can review feedback, revise your answers, and submit again.

1. Offsets 0,1,2,3 were fetched, but only 0 and 2 completed. If commits store the next safe offset, what may be committed?

2. A consumer increments an external counter then crashes before committing its offset. What can happen after restart?

Select one answer.

The broker automatically reverses the first increment The event is guaranteed never to return Exactly-once follows from fetching in order The increment may occur twice unless the effect is deduplicated or transactional

3. A token bucket holds 4 tokens, refills 2 tokens/s, and is emptied at t=0. With no intervening requests, how many tokens are available at t=1.5 seconds?

4. A bounded queue is full and arrival rate persistently exceeds service rate. What helps bound resource usage?

Select one answer.

Apply admission control or backpressure with an explicit rejection policy Retry every rejected job immediately without limit Make the queue unbounded Hide queueing latency from metrics

5. A stream transaction atomically commits output records and input offsets. What remains outside that guarantee unless integrated?

Select one answer.

The committed offsets in that same transaction An arbitrary external HTTP payment side effect The transaction abort state The output records in that same transaction

Auto-advance on pass

Open access learning · Go · Python · C · SQL · No external dependencies

Transparency · Support this project

Join Discord