Module 2 exercises

Start the lab first (docker compose up -d --wait in the code/ folder). Each exercise has a start notebook to work in and a solution notebook to compare with afterwards.

Exercise 1: A key for driver notifications

Goal: choose a record key from an ordering requirement, and prove it works. · Time: ~10 min · Starter: code/module-02/exercise-1/start/

The driver app sends four notifications per charging session: charging started, 80% reached, charging finished and receipt ready. A driver who sees "receipt ready" before "charging started" calls support. The start notebook generates 1,200 notifications for 299 drivers and writes them to the 6-partition topic m02e1-notifications with the key your function key_for(note) returns.

Complete key_for(), run the notebook, and in a markdown cell explain why you didn't choose the session id or the charger id.

You're done when: the check cell prints drivers who saw notifications out of order: 0 of 299 and the busiest partition holds at most twice as many records as the quietest.

Hints

Hint 1 Ask "whose notifications must stay in order?". The answer is the key.
Hint 2 A driver who charges twice in the evening has two sessions. Their notifications must still be in order across both sessions.

Exercise 2: Tune the replay producer

Goal: pick batching and compression settings from measurements. · Time: ~15 min · Starter: code/module-02/exercise-2/start/

After an outage the gateway replays its buffered readings: the same 100,000-reading backlog as in lesson 2.3. The start notebook measures the producer defaults first, then runs each compression codec once. Choose settings for my_settings that, compared with the defaults:

Then write down which codec you'd run in production and why, using your own numbers.

You're done when: all three checks print True, and your note names the codec and the trade-off you accepted.

Hints

Hint 1 Look at the "stored MB" column first: which codecs halve it?
Hint 2 Throughput varies by 10–20% between runs on a laptop. If two codecs are close, decide on size and CPU cost instead. Run the check twice if a result is borderline.