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.
A retail company operates a 12-TB transactional MySQL database that handles high-volume order processing. The data engineering team must replicate this data into BigQuery for near real-time analytics while meeting the following constraints:
Which synchronization workflow should the team implement?
Datastream can write change event streams as self-describing Avro or JSON files in Cloud Storage, which a streaming Dataflow pipeline can consume for custom transformations, enrichment, or loading into downstream targets such as Cloud SQL, Spanner, or BigQuery.
uuid field for deduplication in Dataflow is a critical design flaw: when a pipeline interruption or retry occurs, Datastream re-reads from the last checkpoint and dynamically generates new UUIDs for the exact same source change events, causing duplicate rows in BigQuery.The database were small enough for built-in backfill, custom Dataflow transformations were required in Cloud Storage, and Dataflow deduplicated events using primary keys and change_sequence_number rather than uuid.
Dataflow is a serverless data processing service built on Apache Beam that supports both batch and streaming pipelines. Its JDBC to BigQuery batch template extracts rows from relational databases over a JDBC connection and loads them into BigQuery on a schedule.
last_modified timestamp. Because hard DELETE statements remove rows from the table, the poller cannot detect deletions unless you modify the source schema to use soft-delete flags or add database triggers.SELECT queries against a high-volume 12-TB transactional MySQL database every 15 minutes adds query load unrelated to customer transactions.> or >= can either miss records or ingest duplicates.The source database did not support transaction-log CDC, hard deletes were never performed (or soft deletes were already implemented), and 15-minute batch latency was acceptable.
Because MySQL natively maintains a binary log, log-based CDC via Datastream is strongly preferred over differential querying for maintaining consistency with minimal impact on the source.
Datastream is a serverless Change Data Capture (CDC) and replication service that reads database transaction logs directly without requiring agents or third-party licenses. When synchronizing very large tables where a built-in backfill is impractical, Datastream supports log-positioned CDC combined with an external batch load into BigQuery.
CREATED state without starting it allows the team to extract historical data using an optimized batch export and load it into BigQuery independently, avoiding long-running backfill pressure on the 12-TB production MySQL instance.Unlike SQL polling, scheduled full reloads, or running a slow built-in backfill on a 12-TB database, combining a batch initial load with log-positioned Datastream CDC and BigQuery Merge mode minimizes source impact while ensuring transactional consistency.
In this approach, full table snapshots are periodically exported to Cloud Storage (in formats such as Avro or Parquet) and loaded into BigQuery, while database triggers on source tables capture intermediate inserts, updates, and deletes into shadow journal tables.
The source system lacked any transaction-log CDC mechanism, transactional write volume was very low, and a hybrid batch-plus-trigger design was explicitly permitted.
Modern log-based CDC tools like Datastream read changes asynchronously from the database's existing binary log, achieving the same capture fidelity without adding triggers or delaying transactional commits.
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