Prepare and test your skills

Prepare and test your skills

Free sample question with answers and detailed explanation below. The Professional Cloud Database Engineer practice bank has an additional 769 practice questions.Try 10 free realistic questions: instant score, no credit card, and no account required.
Your team manages an automated nightly workflow using Cloud Scheduler and Cloud Run to export hundreds of BigQuery tables to Cloud Storage to meet a 24-hour Recovery Point Objective (RPO). Recently, transient API quota errors and intermittent network failures caused several table exports to fail silently, while slow-running exports delayed the overall pipeline.
You need to redesign the export monitoring and error-handling strategy to:
Which strategy should you implement?
Using Cloud Scheduler retry settings triggers a full re-execution of the scheduled endpoint if the initial job launch returns a failure status, while a Cloud Monitoring dashboard provides visual charts of job metrics for human operators to inspect manually.
503 service error at launch). If table #150 fails mid-run, the scheduler either remains unaware of the partial failure or restarts the entire export for all tables, duplicating work and wasting quota.The question specifically asked how to recover from transient job-launch failures (like HTTP 503 responses when triggering a pipeline) and how to visualize historical throughput trends alongside an existing automated alerting system.
A sequential monolithic Cloud Run job iterates through a list of tables one by one inside a single execution context. Configuring infinite retries forces the job to keep retrying every failed API response regardless of the HTTP status code, while relying on a weekly manual review of exported logs delays operational oversight.
4xx errors (such as 403 Forbidden or 404 Not Found) indefinitely wastes compute resources and blocks pipeline progress because client configuration or permission errors cannot resolve themselves through retries.You were testing a very small, non-critical development dataset of a few tables where recovery objectives were measured in weeks rather than 24 hours, and you explicitly configured bounded truncated exponential backoff only for transient 5xx or quota errors.
This approach uses a time-triggered Cloud Function to periodically scan Cloud Logging entries every 6 hours and re-trigger the entire export pipeline whenever permission (403) or not-found (404) error codes appear.
403 (Forbidden) and 404 (Not Found) are persistent 4xx client and configuration errors, not transient faults. Automatically resubmitting the workflow will repeatedly fail with the exact same error until an administrator fixes the IAM permissions or corrects the destination bucket path.5xx or quota throttling errors during execution.5xx and quota errors: By focusing retries on 403 and 404 codes, the strategy ignores the transient server and rate-limit errors that actually benefit from automated retries.403/404 errors provides no insight into slow stages or export duration degradation.There is no scenario where automatically retrying 403 or 404 errors indefinitely without human remediation is recommended; however, querying logs or metadata to reconcile and retry only missing records can be useful when an operation lacks built-in per-item tracking and failed due to transient issues.
An event-driven, decoupled export architecture separates table discovery from individual table export execution using Pub/Sub messaging and granular Cloud Run services. Instead of executing one monolithic script, a dispatcher generates one message per table, enabling parallel processing, isolated error handling, and fine-grained observability across Cloud Logging and Cloud Monitoring.
5xx server errors, or temporary network drops—using truncated exponential backoff with jitter, while fatal 4xx configuration errors are logged and acknowledged or sent to a dead-letter topic without infinite looping.This pattern combines asynchronous decoupling, differentiated error handling (retrying transient issues while halting fatal ones), and automated observability. Compared to monolithic batch runs, manual log checks, or passive dashboards, it directly prevents silent failures and keeps large-scale scheduled exports reliable and auditable.
Answer them yourself and get your score the moment you finish. No account needed.
Full access includes
Keep the momentum going with these hand-picked practice scenarios
Last updated Validated