Google ACE Practice Question: Ensuring Successful Operation of a Cloud Solution

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  Create a log-based metric and export the metric to BigQuery

  Why it's wrong here

  Log-based metrics only aggregate matching log entries into counter or distribution metrics; they produce numeric time series, not raw log data. Exporting the metric to BigQuery would give you aggregate counts or value distributions, not the underlying log entries and their structured payloads. Thus this approach cannot satisfy the requirement to export a subset of logs themselves.

• ✓

  Create a log sink with a filter and destination BigQuery

  Why this is correct

  A log sink with a filter and a BigQuery destination is the fully managed, native way to export logs: Cloud Logging continuously routes any newly ingested log entries that match the filter into a specified BigQuery dataset. The sink automatically creates a table with the log schema, and you can use the _PARTITIONTIME pseudo-column for time-based partitioning. This gives reliable, near-real-time export without custom code or manual intervention.

• ✗

  Set up a Cloud Function that triggers on logs and inserts into BigQuery

  Why it's wrong here

  A Cloud Function triggered by logs and inserting into BigQuery lacks the managed, schema-aware routing and buffering that a log sink provides, failing to meet the requirement for a reliable, long-term export of a defined subset. This approach is tempting because it offers custom processing logic for real-time log enrichment or alerting, and would be correct for scenarios requiring immediate, event-driven transformations before storage.

• ✗

  Use gcloud logging read and pipe to bq load

  Why it's wrong here

  Running `gcloud logging read` to fetch logs and piping the output into `bq load` is a point-in-time, manual process that is not suitable for an ongoing, continuous export. It would require your own scheduling, pagination, and error handling, and it consumes time and compute while risking data loss with large log volumes. Additionally, it cannot stream new logs as they arrive, so it actually fails the core requirement of exporting a defined subset in a managed way.

Go deeper

Related to this question