Free Google Professional Cloud Database Engineer Actual Exam Questions

It’s A because using deletion_protection is a clear way to tell Terraform not to destroy the database resource. D is just a fallback, not prevention. B and C don’t address the actual problem of Terraform deleting the DB.

It’s A for sure. Setting deletion_protection to true is the straightforward way to stop Terraform from destroying the database resource during apply, especially when the rest of the environment is being recreated. D sounds like a backup plan rather than prevention, and C doesn’t prevent deletion either. B just reapplies the same config and won’t help if the DB is set to be destroyed. Plus, in Terraform, deletion_protection is a known argument for Cloud SQL instances, so it should work as expected here.

B Cloud Spanner is the only one designed for global transactions with strong ACID guarantees and high uptime, so it’s really the only option that checks all boxes here. The others fall short on either global ACID or availability.

Cloud SQL might seem like a solid choice because of its strong ACID support, but it’s not really built for that 99.999% global availability or scale. Firestore does offer some ACID transactions but it’s more tailored for semi-structured data and not full relational needs. Bigtable just can’t guarantee ACID at the transaction level, so it’s out. Cloud Spanner stands out because it’s basically built exactly for this kind of global, strongly consistent transactional workload. Anyone else think Firestore could still be useful if the transactional scope was smaller?

I think B is the better pick here. Multi-region Cloud Spanner is designed for global scale and continuous availability, which fits the 24/7 minimal downtime requirement way better than regional setups or Cloud SQL options. Starting with A might save cost initially, but you'd have to switch later, adding complexity. C and D just don’t handle the global scale and uptime as smoothly. So, B makes the most sense from the start to avoid downtime and support millions of users worldwide.

C/D? Cloud SQL can scale but might struggle with global 24/7 uptime.

B imo, adding more replicas definitely helps distribute the reporting load better. A also makes sense since indexes speed up query execution without changing reports. E feels like a bigger shift than the question might want.

B – More replicas spread the load and reduce report delays. A – Secondary indexes speed up queries without changing

Option E seems like a solid choice since pglogical is critical for logical replication in DMS migrations. For the second pick, I’d go with B. Disabling foreign key constraints on the source database can help avoid issues with data consistency and ordering during the migration. C is important but usually more for change data capture scenarios, and shutting down the database (D) is definitely not needed for DMS since it supports online migrations. So E and B make the most sense to me given this context.

E imo, pglogical is usually needed for logical replication with DMS, so that one makes sense. But I’m not sure about the primary key requirement—C might be important because some tools need it for replication or change tracking? Also, the question doesn’t specify if this is a one-time migration or ongoing replication, which could change what’s required. Does anyone know if Cloud SQL supports migrations without shutting down the source DB? That would make D less likely.

Option A also makes sense since scheduling maintenance during off-hours can minimize user impact, but it doesn't fully prevent downtime like HA does. It’s more about timing than avoiding downtime altogether.

D. High availability is designed to keep the instance available during maintenance by failing over to a standby, so no downtime should occur. B with a read replica helps with read scaling but doesn’t prevent downtime for writes during maintenance on the primary. The question stresses no downtime, not just read availability, so HA is more aligned with that need. Maintenance windows (A) or notifications (C) don’t eliminate downtime, just try to minimize user impact. So, enabling HA seems like the most straightforward way to ensure continuous availability during updates.

It’s A. Using Cloud Scheduler with Cloud Functions is super lightweight and avoids the cost and hassle of managing a VM or Composer environment. Plus, it fits well for simple scheduled exports.

Maybe D could work since it keeps the existing cron job logic intact without extra refactoring, but it’s definitely more expensive and needs more maintenance than serverless options.

Makes sense to avoid the dump import since it’s slow and causes downtime. Option D sounds like the best fit because creating an external replica lets you keep data synced continuously without heavy maintenance. Plus, it matches Google’s recommended approach for minimal downtime migrations from on-prem MySQL to Cloud SQL.

D, external replica syncs live data, unlike C’s slow dump import.

C imo, Bigtable is built for high-volume, low-latency time-series data like this.

C/D? Bigtable is perfect for time-series data at high throughput, but if you need SQL and strong consistency across regions, Spanner fits better. Since the question highlights real-time monitoring, Bigtable's speed matters more here.

D, because custom services (A) add complexity and federated queries (B) aren’t real-time.

C/D? Database Migration Service (C) is usually for one-time migrations, not continuous replication. Datastream with Dataflow (D) fits near-real-time needs better and is fully managed, so D still seems more solid here.

B/D? PgBouncer (B) is a solid pick since it’s built for connection pooling and can help route read queries to replicas, which matches the load-splitting goal. The Cloud SQL Auth proxy (D) mainly handles secure connections and authentication, not really load balancing or pooling between instances. So even though D is useful for security, it doesn’t directly address splitting the load. A and C are definitely out because they’re meant for web traffic, not database connections.

This one’s tricky, but I’d rule out A and C since Cloud Load Balancing is for HTTP/HTTPS traffic, not database connections. Between B and D, PgBouncer (B) is specifically designed for connection pooling and can be configured to route read queries to replicas, making it a good fit for splitting loads in a database setup. The Cloud SQL Auth proxy (D) mainly handles secure connections, not load distribution. So I’d go with B here.

Maybe D because PostgreSQL's native authentication can be set up with SSL certificates for password-less access, which works directly without relying on external identity providers.

Option A seems solid because IAM lets you use tokens instead of passwords, which fits password-less needs for both read and write without relying on external directories.

B/C? B helps spot if the problem’s resource-related, but C might catch repeated errors that aren’t obvious in logs or metrics. Both could give clues from different angles.

Maybe A is better here since Logs Explorer can reveal slow queries or specific errors causing the drop in performance, not just general resource issues. That’s more targeted for root cause analysis.

Makes sense to start with A here since increasing processing units directly tackles CPU load before considering more complex solutions like sharding or schema tweaks.

A. Before jumping into schema changes or sharding, Google docs suggest upping processing units since it’s the fastest way to relieve CPU pressure in Cloud Spanner.

B/D? Cloud Data Fusion is managed and easier to set up with scheduled workflows, though maybe not as low latency as streaming. D adds complexity with replication and federated tables, which might not be as straightforward.

C seems best for low latency and managed streaming from on-prem MySQL to BigQuery, using Datastream and Dataflow is pretty straightforward. The others feel more batch or complex.