Free PCDBE sample questions

Correct answer: In the Cloud SQL instance settings, set the `cloudsql.enable_pgaudit` database flag to `on` and configure the `pgaudit.log` flag to include `READ` operations.

Standard Cloud Audit Logs for Cloud SQL do not capture the full SQL text for data access events like SELECT queries for performance reasons. To meet the strict requirement of logging the exact SQL text for administrative queries, you must enable the `pgaudit` extension. This is done by setting the `cloudsql.enable_pgaudit` flag to `on` and then configuring the `pgaudit.log` flag to specify which classes of commands to log, such as `READ` for select statements.

Correct answer: The staleness value is too low, forcing the query to wait for a timestamp that has not yet been propagated to the nearest replica.

Cloud Spanner serves stale reads from the closest replica that has data up to the requested timestamp. If the staleness bound (e.g., `max_staleness` of 1 second) is too aggressive, the nearest replica may not have received the data for that timestamp yet. In this case, the query must wait for the data to propagate, which can introduce latency and make the stale read perform similarly to a strong read that goes to the leader region.

Correct answers: Configure a continuous DMS migration job to keep the Cloud SQL target synchronized with the on-premises Oracle source.; Establish a reverse replication stream from the Cloud SQL target back to the on-premises Oracle source after the cutover.

A continuous migration job using Change Data Capture (CDC) is essential for minimizing downtime. It performs the initial bulk copy and then streams ongoing changes, allowing the target database to remain in sync with the source until the final cutover.

Setting up reverse replication is a critical part of a fallback plan. After promoting the Cloud SQL instance and directing traffic to it, this reverse stream sends any new changes back to the original source database. This keeps the source 'warm' and enables a rapid rollback to the on-premises environment if a critical issue is discovered in the new Google Cloud application.

Correct answer: Cloud Bigtable, with a row key of `userId#reversed_timestamp`. This design prevents hotspotting and keeps a user's events contiguous and sorted for efficient reads.

Cloud Bigtable is ideal for massive-scale, low-latency write workloads like this event ingestion scenario. The key to performance is the row key design. A `userId` prefix groups all events for a single user together. Appending a reversed timestamp (e.g., `Long.MAX_VALUE - timestamp`) ensures that the most recent events are stored at the beginning of the row range for that user. This makes the query for the 'last N events' a highly efficient scan of just a few rows, satisfying all requirements.

Correct answer: False

This statement is false. The purpose of the standby instance in a Cloud SQL HA configuration is solely for failover. It is a 'hot standby' that takes over if the primary instance becomes unavailable. It does not serve read traffic. To scale read throughput, you must create one or more separate read replicas.

Correct answer: Configure a cross-region read replica and have a documented procedure to manually promote it in case of a regional outage.

The standard HA configuration for Cloud SQL protects against zonal failures, not regional failures. To meet a low RTO for a regional disaster, you must have a replica in another region. A cross-region read replica serves this purpose. While replication is asynchronous (meaning RPO is not strictly zero, but very close), it is the standard and required Cloud SQL pattern for regional DR. The promotion must be handled manually or via automation scripts, as it's not an automatic process like zonal HA failover.

Correct answer: Create a composite index on the `(category, is_active)` columns of the `products` table.

The problem describes queries filtering by two specific columns, which are causing full table scans. This is a classic indicator of a missing index. A composite index on the columns used in the `WHERE` clause (`category` and `is_active`) will allow the database to directly look up the required rows instead of scanning the entire table, which is the most direct and efficient solution.

Correct answer: A composite index on the `inventory` array field and the `skill_level` field.

Firestore requires an index for every query it executes. A simple query on one field will use a default single-field index. However, this query involves a range filter (`skill_level > 90`) and an array membership filter (`inventory` contains `rare_item`). To perform this query efficiently, Firestore requires a manually created composite index that includes both fields. Without it, the query would be rejected by the Firestore API.

Correct answer: 1. Configure source connectivity. 2. Create and run a continuous migration job. 3. Monitor replication delay. 4. Stop the source application. 5. Promote the Cloud SQL replica. 6. Reconfigure the application to connect to Cloud SQL.

This sequence follows the Google-recommended best practices for a near-zero downtime migration with DMS. You first set up connectivity, then start the continuous job which handles the initial dump and ongoing replication. It's crucial to monitor the replication delay to ensure the target is caught up before cutover. The cutover itself involves stopping the application to prevent new writes to the old source, promoting the Cloud SQL instance to a standalone, writable server, and finally, pointing the application to the new database.