Google Cloud Professional Cloud Database Engineer Practice Exam

What the Google Cloud Professional Cloud Database Engineer exam covers

• Design innovative, scalable, and highly available cloud database solutions243 questions
• Manage a solution that can span multiple database technologies186 questions
• Migrate data solutions174 questions
• Deploy scalable and highly available databases in Google Cloud151 questions

Free Google Cloud Professional Cloud Database Engineer practice test questions

A sample of 10 questions with answers and explanations. Sign up free to practice all 754.

1. Question 1Design innovative, scalable, and highly available cloud database solutions

   A global payments company is building a ledger that must serve reads and writes from users on three continents, guarantee strong external consistency across all regions, and scale horizontally to millions of transactions per second without any sharding logic in the application. Which Google Cloud database best fits these requirements?

   • ACloud SQL for PostgreSQL with cross-region read replicas
   • BCloud Spanner in a multi-region instance configurationCorrect
   • CFirestore in Native mode with composite indexes
   • DBigtable with multi-cluster routing enabled
   ✓ Correct answer: B

   A financial ledger needs ACID transactions, relational semantics, and correctness across continents without the application managing shards. Cloud Spanner's TrueTime-based external consistency and automatic sharding deliver exactly this at scale, which no other listed service provides simultaneously.

   Why the other options are wrong

   • ARead replicas are asynchronous, so they give neither global write scale nor external consistency.
   • CFirestore is a document store and does not offer relational SQL transactions at that write scale.
   • DBigtable is NoSQL wide-column and has no cross-region multi-row ACID transactions.
2. Question 2Design innovative, scalable, and highly available cloud database solutions

   A Cloud SQL for MySQL instance shows disk I/O saturation during peak writes, yet CPU and memory sit well below their limits. What is the most effective way to raise sustained IOPS and throughput?

   • AUpgrade to a machine type with more vCPUs and memory.
   • BIncrease the instance storage capacity, since IOPS scale with disk size.Correct
   • CConvert the data disk from SSD to HDD storage.
   • DAdd a read replica to absorb the write I/O load.
   ✓ Correct answer: B

   Cloud SQL ties SSD IOPS and throughput to the size of the disk, so an I/O-bound instance with spare CPU and memory benefits from a larger disk rather than a bigger machine. This lifts sustained write performance without changing the engine.

   Why the other options are wrong

   • ACPU and memory are not the bottleneck here.
   • CHDD provides fewer IOPS than SSD.
   • DRead replicas offload reads, not write I/O.
3. Question 3Design innovative, scalable, and highly available cloud database solutionsSelect all that apply

   An architect is designing a Cloud Spanner schema for a high-throughput ledger and wants to avoid write hotspots and unnecessary latency. Which TWO practices should they apply? (Select TWO.)

   • AUse a single sequential integer key so rows insert in order
   • BAvoid monotonically increasing values as the leading key columnCorrect
   • CInterleave child rows with their parent to keep related data togetherCorrect
   • DAdd every column to a secondary index to speed all queries
   • EPlace the newest timestamp first in the key to group recent writes
   ✓ Correct answer: B, C

   Spanner shards by key range, so high-entropy leading keys plus interleaving of tightly related entities give both even write distribution and low-latency locality.

   Why the other options are wrong

   • AA sequential key is exactly what causes the write hotspot to be avoided.
   • DIndexing every column adds write cost and does not prevent hotspots.
   • EA leading timestamp concentrates all new writes on the newest split.
4. Question 4Design innovative, scalable, and highly available cloud database solutionsSelect all that apply

   A security team is writing runbooks for CMEK-protected databases. Which two statements are true about permanently destroying a Cloud KMS key version, as opposed to merely disabling it? (Select TWO.)

   • AData protected only by that version becomes unrecoverableCorrect
   • BData re-encrypts under Google-managed keys automatically
   • CA destroyed version can be restored like a disabled one
   • DDestruction is scheduled with a mandatory waiting periodCorrect
   • EDestroying the version speeds up query performance
   ✓ Correct answer: A, D

   Cloud KMS key destruction is a two-step process: the version first enters a scheduled-for-destruction state for a configurable waiting period (24 hours by default) during which it can still be restored, and after that it is destroyed permanently. Once destroyed, any data whose DEK was wrapped solely by that version can no longer be decrypted. There is no automatic fallback to Google-managed keys.

   Why the other options are wrong

   • BData does not silently fall back to Google-managed keys.
   • CAfter the waiting period a destroyed version cannot be restored.
   • EKey destruction has no effect on query performance.
5. Question 5Manage a solution that can span multiple database technologiesSelect all that apply

   An SRE team monitoring a production Cloud SQL fleet must document the uptime SLA and track compliance against it. Which TWO statements accurately describe Google's Cloud SQL uptime SLA and how you monitor against it? (Select TWO.)

   • AOnly high-availability (regional) instances are covered by the uptime SLACorrect
   • BYou can track measured uptime and error budget using Cloud Monitoring SLOsCorrect
   • CThe SLA guarantees no maintenance restarts ever occur
   • DRead replicas are covered by the same uptime SLA as the primary
   • ESLA credits are applied automatically with no customer action
   ✓ Correct answer: A, B

   Only high-availability configurations qualify as covered instances under the SLA, and Cloud Monitoring SLOs let you track uptime and error-budget burn against your target. The SLA does not promise zero maintenance, does not extend to replicas the same way, and credits must be requested.

   Why the other options are wrong

   • CMaintenance can still cause restarts; the SLA is an availability target, not a no-restart promise.
   • DRead replicas are generally excluded from the primary's uptime SLA coverage.
   • ESLA credits require the customer to submit a claim, not automatic application.
6. Question 6Manage a solution that can span multiple database technologiesSelect all that apply

   A DBA is documenting the correct uses of Cloud SQL read replicas for the team. Which TWO statements about Cloud SQL read replicas are accurate? (Select TWO.)

   • AThey serve read-only queries to offload the primaryCorrect
   • BThey can be located in a different region from the primaryCorrect
   • CThey accept writes that then sync back to the primary
   • DThey provide automatic failover for high availability
   • EThey share a single storage volume with the primary
   ✓ Correct answer: A, B

   Cloud SQL read replicas replicate asynchronously and serve reads only. Cross-region replicas add geographic read locality and DR value. They are distinct from the HA standby, which handles failover.

   Why the other options are wrong

   • CReplicas are read-only and never write back to the primary.
   • DAutomatic failover requires an HA configuration, not read replicas.
   • EEach replica has its own independent storage, not a shared volume.
7. Question 7Migrate data solutions

   An organization is replatforming a SQL Server application onto Cloud SQL for PostgreSQL. The application relies heavily on T-SQL stored procedures and SQL Server specific data types. Which statement correctly describes what this migration must include that a same-engine migration would not?

   • AThe schema and procedural code must be converted to PostgreSQL equivalents before data loadsCorrect
   • BThe destination must run in the same region as the source to preserve collation
   • CThe migration can only proceed during an extended maintenance outage
   • DContinuous change data capture is unavailable for any PostgreSQL destination
   ✓ Correct answer: A

   Moving from SQL Server to PostgreSQL changes the engine, so T-SQL stored procedures and SQL Server data types have no automatic equivalent and must be converted first. A same-engine (homogeneous) migration keeps the identical schema and needs no such translation. This conversion step is the defining extra work of a heterogeneous migration.

   Why the other options are wrong

   • BCo-locating regions does not preserve SQL Server collation; collation is handled during schema conversion.
   • CHeterogeneous migrations still support CDC-based near-zero-downtime cutover, so an extended outage is not mandatory.
   • DDMS supports CDC into Cloud SQL for PostgreSQL and AlloyDB, so change data capture is available.
8. Question 8Migrate data solutions

   A company is moving from Oracle to AlloyDB and will redesign the schema and rewrite parts of the application to use PostgreSQL-native partitioning and features rather than preserving the original design. Which migration approach does this describe?

   • ARefactor and re-architect, redesigning the schema and application for the target platformCorrect
   • BLift-and-shift, relocating the workload unchanged onto an equivalent managed database instance
   • CReplatform, keeping the application the same while swapping only the managed engine
   • DRepurchase, replacing the system with a different commercial SaaS product
   ✓ Correct answer: A

   Refactoring or re-architecting means substantially changing the schema and application to take advantage of target-native capabilities, trading higher effort for better fit and performance. Rewriting the app around PostgreSQL partitioning and features is exactly this pattern, as opposed to moving the workload unchanged.

   Why the other options are wrong

   • BLift-and-shift moves the workload as-is with minimal change, which contradicts the schema redesign and rewrite described.
   • CReplatform makes only modest changes, such as swapping the engine, without rearchitecting the application logic.
   • DRepurchase replaces the system with a different off-the-shelf product, not a redesign onto AlloyDB.
9. Question 9Deploy scalable and highly available databases in Google Cloud

   A Cloud SQL for PostgreSQL production instance on the Enterprise edition cannot tolerate the roughly minute-long restart that maintenance updates cause. The team needs maintenance applied with sub-second interruption. What should they do?

   • ASet a deny maintenance period so the Enterprise-edition maintenance never restarts the instance.
   • BAdd a read replica so the primary can be taken offline for maintenance without user impact.
   • CMove the instance to the Enterprise Plus edition, which applies maintenance with near-zero (sub-second) downtime.Correct
   • DConfigure a maintenance window at 03:00, which changes maintenance to a rolling, zero-downtime process.
   ✓ Correct answer: C

   The Enterprise Plus edition applies maintenance updates with near-zero downtime (typically under a second) by using an orchestrated failover, whereas the Enterprise edition incurs a restart on the order of a minute. Upgrading the edition is the supported way to meet a sub-second interruption target for planned maintenance.

   Why the other options are wrong

   • AA deny maintenance period only defers maintenance up to a limit; it does not eliminate the eventual restart.
   • BA read replica does not make the primary's maintenance transparent; writes still land on the primary during its restart.
   • DA maintenance window controls when maintenance runs, not whether it restarts the instance; Enterprise-edition maintenance still causes downtime.
10. Question 10Deploy scalable and highly available databases in Google Cloud

    A serverless application scales to thousands of concurrent instances, and the Cloud SQL for PostgreSQL primary is exhausting its max_connections limit during peaks. The team needs to manage connections at scale without over-provisioning the database. What should they implement?

    • AIncrease max_connections to the maximum and scale up the instance to match.
    • BGive each application instance its own cross-region read replica to connect to.
    • CPut a connection pooler in front of the database to multiplex client connections.Correct
    • DSwitch all clients to the public IP endpoint to lower per-connection overhead.
    ✓ Correct answer: C

    A pooler such as PgBouncer or Cloud SQL managed connection pooling sits between the app and the database and shares a small set of backend connections across thousands of clients. This keeps the database within its connection limits without scaling it up purely for connection count. It is the standard pattern for high-fan-out serverless workloads.

    Why the other options are wrong

    • ARaising max_connections and scaling up wastes resources and still hits a ceiling as instances keep growing.
    • BA replica per app instance multiplies infrastructure and does not reduce total connections to the primary.
    • DThe public IP endpoint changes the network path, not the number of connections the database must sustain.

Who this Google Cloud Professional Cloud Database Engineer practice exam is for

This practice set is for anyone preparing for the Google Cloud Professional Cloud Database Engineer exam at the advanced level - from first-time candidates building a foundation to experienced Google practitioners doing a final review before test day. If you learn best by working through realistic questions and reading why each answer is right or wrong, it is built for you.

How to use this Google Cloud Professional Cloud Database Engineer practice exam

1. Start with the free sample questions above to gauge your current baseline.
2. Read the full explanation on every question, including why each wrong option is wrong.
3. Track your weak domains and focus your study where you are losing the most marks.
4. Once you are scoring consistently well, take a timed, full-length mock exam.
5. Use your readiness score to decide when you are ready to book the real Google Cloud Professional Cloud Database Engineer exam.

Related Google resources

• Google Cloud Professional Cloud Database Engineer study guideKey concepts
• Google practice examsAll Google
• Certification pathWhere this fits
• Certification exam guides & tipsBlog
• Plans & pricingFree & paid
• How these questions are written and reviewedMethodology
• Report a problem with a questionCorrections
• Google Cloud Professional Cloud Developer practice examRelated
• Google Cloud Professional Cloud DevOps Engineer practice examRelated
• Google Cloud Professional Cloud Network Engineer practice examRelated

Google Cloud Professional Cloud Database Engineer practice exam FAQ