GCP Professional Data Engineer Practice Question

Your online gaming platform sends player action events to Pub/Sub, then processes them in an Apache Beam/Dataflow pipeline. A game session is a burst of events separated by less than 20 minutes of inactivity. The pipeline must: emit the session result as soon as the gap closes; accept late events for up to 5 minutes and issue one update; drop anything later. Which Beam windowing and trigger setup meets these requirements?

Use 1-hour fixed (tumbling) windows with the default AfterWatermark trigger and no allowed-lateness period.
Place all events in a global window and register processing-time timers to close sessions after 20 minutes of inactivity, disabling allowed lateness.
Configure a session window with a 20-minute gap duration, set 5-minute allowed lateness, and use an AfterWatermark trigger with one late firing in accumulating mode.
Apply 20-minute sliding windows with a 5-minute slide interval and AfterProcessingTime triggers firing every minute.

Report Issue

Answer Description

A session window groups events that are separated by no more than a specified gap duration, which matches the 20-minute inactivity rule. Setting an allowed-lateness of 5 minutes tells Dataflow to keep the window pane open for that period to accept straggling events. Using an AfterWatermark.pastEndOfWindow() trigger with a single late firing-such as withLateFirings(AfterProcessingTime.pastFirstElementInPane(Duration.standardMinutes(5)))-and ACCUMULATING mode ensures that:

The first pane is emitted as soon as the watermark passes the session end (requirement a).
Exactly one additional pane is produced if late data arrive within 5 minutes, updating the previous result (requirement b).
Events arriving after the allowed-lateness window are dropped (requirement c). Fixed or sliding windows cannot model user-driven session boundaries, and a global window would require custom gap-detection logic, increasing complexity.

Ask Bash

Bash is our AI bot, trained to help you pass your exam. AI Generated Content may display inaccurate information, always double-check anything important.