Google Cloud Professional Cloud Developer Practice Exam

What the Google Cloud Professional Cloud Developer exam covers

• Designing Cloud-Native Applications259 questions
• Building and Testing Applications206 questions
• Deploying Applications150 questions
• Integrating Google Cloud Services173 questions

Free Google Cloud Professional Cloud Developer practice test questions

A sample of 10 questions with answers and explanations. Sign up free to practice all 788.

1. Question 1Designing Cloud-Native Applications

   You are building a web application that runs on multiple instances behind a load balancer and scales horizontally. Where should user session state be kept so that any instance can serve any user's request?

   • AIn a shared external store such as Memorystore for Redis that every instance can accessCorrect
   • BIn each instance's local process memory, which new instances inherit automatically on scale out
   • CIn each instance's local memory, relying on the autoscaler to replicate it to new instances
   • DIn local memory on the first instance, which then pushes it to instances added later
   ✓ Correct answer: A

   Each instance runs its own isolated process memory, so an instance added during a scale out event cannot see session data held only in another instance. Externalizing session state to a shared store such as Memorystore for Redis lets any instance serve any request and keeps the application layer stateless.

   Why the other options are wrong

   • BProcess memory is not shared between instances, so new instances start empty rather than inheriting state.
   • CAutoscalers do not replicate process memory between instances.
   • DThere is no mechanism that pushes one instance's in memory state to instances added later.
2. Question 2Designing Cloud-Native Applications

   Developers at Driftwood Logistics currently hand write client libraries in several languages to call their REST API, which is slow and error prone whenever the API changes. What artifact would let them generate client libraries automatically instead of hand writing them?

   • AA cron schedule expression describing job run timing
   • BA Terraform configuration describing the infrastructure
   • CA Dockerfile describing how to build the container image
   • DAn OpenAPI specification describing paths and schemasCorrect
   ✓ Correct answer: D

   Tools that read an OpenAPI document can generate typed client code in many languages directly from the described paths and schemas, so updates to the document can regenerate the libraries instead of hand editing them. A Terraform configuration describes cloud infrastructure resources, not an API's request and response shapes. A Dockerfile describes how to build a container image, and a cron schedule describes job timing, neither of which relate to describing an API contract for client generation.

   Why the other options are wrong

   • AA cron schedule expression only describes timing for a scheduled job and carries no information about an API's structure.
   • BA Terraform configuration provisions infrastructure resources and does not describe an API's request and response schema needed for client code generation.
   • CA Dockerfile defines how to build a container image and has no information about an API's paths or data schemas.
3. Question 3Designing Cloud-Native Applications

   A data engineering team notices that most BigQuery queries filter on an order date column, but every query scans the entire multi year table and cost has become high. What should they do to reduce cost?

   • APartition the table by the order date columnCorrect
   • BAdd a secondary index on the order date column
   • CEnable a Memorystore cache in front of BigQuery
   • DMove the table into Cloud SQL instead
   ✓ Correct answer: A

   Once the table is partitioned by order date, queries that filter on that column only scan the relevant partitions instead of the whole multi year table, directly reducing bytes scanned and cost. BigQuery does not use traditional secondary indexes for this, moving years of data into Cloud SQL would hurt analytical performance, and Memorystore caching does not apply to BigQuery's scan based execution model.

   Why the other options are wrong

   • BBigQuery does not use traditional secondary indexes to prune scanned data; partitioning is the mechanism for that.
   • CMemorystore caching does not apply to BigQuery's query execution model for large table scans.
   • DMoving years of analytical data into Cloud SQL would hurt performance and is not built for this scale of analytics.
4. Question 4Designing Cloud-Native Applications

   A security team is worried that individual project owners keep generating service account keys despite repeated reminders to use attached identities instead. What organization level control would prevent any new key from being created across all projects?

   • AThe iam.disableServiceAccountKeyCreation organization policy constraintCorrect
   • BAn IAM deny policy that only blocks key creation for one named project
   • CA Cloud Audit Log alert that emails the security team after each key is made
   • DA Cloud Function that automatically deletes any key found during a scan
   ✓ Correct answer: A

   Setting this constraint at the organization or folder level prevents the action from succeeding in the first place, rather than only detecting or cleaning up keys after they are created. This is a stronger preventive control than logging, alerting, or after the fact deletion.

   Why the other options are wrong

   • BScoping the deny policy to a single project leaves every other project still able to create keys.
   • CAn alert only notifies the team after a key already exists, it does not prevent the key from being created.
   • DA nightly cleanup function still allows a window where a freshly created key is valid before it gets deleted.
5. Question 5Building and Testing Applications

   A developer bypasses the client library and calls the Cloud Storage JSON API directly over HTTP to list objects in a bucket holding millions of objects. The first response contains only 1,000 object entries. What must her code do to retrieve the remaining objects?

   • ARepeat the exact same first request over and over until the returned count matches expectations
   • BIncrease the maxResults parameter until the whole bucket fits in a single response
   • CRead the nextPageToken value and pass it as pageToken on the following requestCorrect
   • DAdd an Accept-Range header to the following request to request a different byte range
   ✓ Correct answer: C

   Unlike a client library's automatic pager, direct REST calls return only one page at a time along with a token identifying the next page, if any. The caller is responsible for looping, passing that token as a request parameter, and stopping once no token is returned.

   Why the other options are wrong

   • ARepeating the exact same request returns the same first page again rather than advancing to subsequent results.
   • BServer side page size limits mean a very large bucket cannot be forced into a single response regardless of the requested maximum.
   • DAccept-Range headers apply to byte range requests on object content, not to paging through a list of object metadata entries.
6. Question 6Building and Testing ApplicationsSelect all that apply

   A team is planning a load and performance test for a newly launched serverless API before a major marketing event. Which three goals should this test aim to achieve? Choose three.

   • AConfirm that the service's API documentation is up to date and grammatically correct
   • BMeasure how cold start delays affect response times under sustained loadCorrect
   • CConfirm that autoscaling limits allow enough instances for peak trafficCorrect
   • DVerify that individual unit tests for business logic still pass under normal load
   • EDetermine the maximum sustainable throughput before errors or latency degradeCorrect
   ✓ Correct answer: B, C, E

   Understanding maximum sustainable throughput, how cold starts affect latency under load, and whether autoscaling configuration allows enough instances to be provisioned are all central goals of performance testing before an expected traffic spike. These insights let a team adjust configuration such as minimum instances or concurrency settings ahead of the event. Verifying unit test correctness or documentation quality are valuable activities but are unrelated to load and performance testing.

   Why the other options are wrong

   • ADocumentation quality is unrelated to how the service performs under load and is not a goal of performance testing.
   • DUnit test correctness is validated earlier in the pipeline and is not something a load test is designed to check.
7. Question 7Deploying Applications

   A container occasionally deadlocks internally so that its process keeps running but stops responding to any requests. Which probe type should the manifest define so GKE automatically restarts the container when this happens?

   • AReadiness probe
   • BResource limit
   • CLiveness probeCorrect
   • DStartup probe
   ✓ Correct answer: C

   Kubelet uses the liveness probe result to decide whether a container is still functioning correctly; repeated failures cause the container to be killed and restarted according to the pod restart policy. Readiness probes instead only control whether traffic is sent to the pod, without restarting it.

   Why the other options are wrong

   • AA readiness probe only removes the pod from Service endpoints when it fails, it does not restart the container.
   • BResource limits cap CPU and memory usage and have no mechanism for detecting or reacting to an unresponsive process.
   • DA startup probe is meant to delay other probes until a slow-starting container finishes initializing, not to catch a later deadlock.
8. Question 8Deploying Applications

   A backend team wants other pods in the same GKE cluster to reach a group of pods using a stable DNS name and virtual IP, while load balancing traffic across all healthy pod replicas. Which resource satisfies this requirement?

   • ANetworkPolicy
   • BConfigMap
   • CIngress
   • DServiceCorrect
   ✓ Correct answer: D

   Kubernetes Services use label selectors to group pods and continuously update their endpoint list as pods come and go, giving internal callers a consistent name to connect to. Ingress instead manages external HTTP routing and typically fronts one or more Services rather than replacing them.

   Why the other options are wrong

   • ANetworkPolicy controls which pods are allowed to communicate with each other, but it does not provide service discovery or load balancing.
   • BConfigMap stores configuration data for pods to consume and has no role in exposing or load balancing network traffic.
   • CIngress configures external HTTP or HTTPS routing rules into the cluster and relies on Services underneath, so it does not itself provide internal cluster DNS load balancing.
9. Question 9Integrating Google Cloud Services

   A team instrumenting a Python microservice with OpenTelemetry wants the spans their application creates to be sent to Cloud Trace. What must they configure in the OpenTelemetry SDK?

   • AA MeterProvider configured with the Cloud Monitoring exporter
   • BA ContextPropagator configured with the B3 format
   • CA TracerProvider configured with the Cloud Trace exporterCorrect
   • DA LoggerProvider configured with the Cloud Logging handler
   ✓ Correct answer: C

   OpenTelemetry separates tracing, logging, and metrics into distinct providers, and spans specifically flow through a TracerProvider, which must be configured with an exporter that knows how to send span data to the Cloud Trace API. A LoggerProvider handles log records, a MeterProvider handles metrics, and a propagator format like B3 controls how context is carried across service boundaries, none of which is the piece that actually ships spans to Cloud Trace.

   Why the other options are wrong

   • AA MeterProvider with a Cloud Monitoring exporter publishes metrics, not the trace spans described here.
   • BA propagator controls how trace context is carried between services, it does not by itself export spans to Cloud Trace.
   • DA LoggerProvider with a Cloud Logging handler ships log records, it has no role in exporting trace spans.
10. Question 10Integrating Google Cloud Services

    A field service app writes a status update to Firestore while offline. The write is queued locally by the SDK's offline persistence. How can the app's listener callback determine that this particular update has not yet been acknowledged by the backend?

    • ACheck the fromCache property, which is set only for pending writes
    • BCheck the exists property on the document snapshot
    • CCheck the hasPendingWrites property on the snapshot metadataCorrect
    • DCompare the document's update time to the current device clock
    ✓ Correct answer: C

    When the offline enabled SDK queues a write, listener callbacks for affected documents fire with local data whose metadata.hasPendingWrites is true. Once the server acknowledges the write, the listener fires again with hasPendingWrites set to false.

    Why the other options are wrong

    • AThe fromCache property indicates whether the snapshot came from the local cache rather than the server, it is not specifically the marker for pending writes.
    • BThe exists property only indicates whether a document is present, it does not reveal whether a write is still pending server confirmation.
    • DComparing update time to the device clock is unreliable and is not how Firestore represents pending write state.

Who this Google Cloud Professional Cloud Developer practice exam is for

This practice set is for anyone preparing for the Google Cloud Professional Cloud Developer 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 Developer 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 Developer exam.

Related Google resources

• Google Cloud Professional Cloud Developer 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 DevOps Engineer practice examRelated
• Google Cloud Professional Cloud Network Engineer practice examRelated
• Google Cloud Professional Cloud Security Engineer practice examRelated

Google Cloud Professional Cloud Developer practice exam FAQ