Google CloudUpdated 2026Last reviewed

Google Professional Cloud Developer (PCD) Practice Test

Build Cloud-Native Apps. Pass Professional Cloud Developer.

500 scenario-based practice questions covering all 4 Professional Cloud Developer exam sections: designing scalable, secure, and reliable cloud-native applications, building and testing, deploying to Cloud Run and GKE, and integrating with Google Cloud services.

No credit card required · Upgrade to 500 questions for $11.99

500 Exam-Style Questions
7-Day Money-Back Guarantee

Is this for you?

You ship code to Cloud Run or GKE and want practice deciding when Workflows, Cloud Tasks, Cloud Scheduler, or Eventarc should coordinate a set of services
You can write the application but want to check how you would authenticate it: Application Default Credentials, a service account, the Cloud SQL Auth Proxy, or Workload Identity Federation
You want to test yourself on API call patterns the guide lists, such as batching requests, restricting return data, paginating results, and retrying with exponential backoff
You need practice reading a symptom and choosing the fix, for example correlating trace spans with trace IDs or using Error Reporting to group application failures

Not for you if:

If your work is mostly provisioning projects, networks, and IAM rather than writing application code, Associate Cloud Engineer covers that ground and is a better first Google exam
If you want leaked real exam questions, we do not offer braindumps; every question here is written against the published PCD exam guide

The Google Professional Cloud Developer (PCD) is a professional-level Google Cloud certification. The exam has 50–60 questions, runs 120 minutes, costs $200, and requires a passing score of Not published by Google. CertSharp provides 500 exam-style PCD practice questions — 30 free with no credit card — each with a detailed explanation.

Start 2–3 weeks before your exam for best results
Updated for 2026 exam blueprint
Detailed explanations for every question

Exam Overview

PCD Exam Details

Know exactly what to expect on exam day

Questions

50–60

Duration

120 minutes

Passing Score

Not published by Google

Exam Cost

$200

Compare this fee with other certifications in our exam cost comparison, or check your readiness with the free PCD readiness quiz.

Coverage

Exam Domains Covered

Every question maps to an official PCD domain

1

Designing highly scalable, secure, and reliable cloud-native applications (~32%)

2

Building and testing applications (~23%)

3

Configuring cloud-native applications for deployment (~24%)

4

Integrating applications with Google Cloud services (~21%)

From the exam guide

What the PCD exam actually tests

PCD is written from the developer's seat: most considerations in the guide are about code you write, build, deploy, and debug, not infrastructure you administer. The largest section (~32%) is design, so many questions ask you to pick a platform, data store, or integration pattern for a stated requirement before any deployment detail comes up.

Picking the runtime and rollout

The guide asks you to choose between Compute Engine, GKE, and Cloud Run for a use case, then plan traffic splitting for gradual rollouts, rollbacks, or A/B tests. On GKE you also need to know how health checks and Horizontal Pod Autoscaler metrics affect availability.

Event-driven and orchestrated services

Expect to decide between Pub/Sub, Eventarc, Cloud Tasks, Cloud Scheduler, and Workflows for connecting services. You should know how a Cloud Run service is invoked by an Eventarc or Pub/Sub trigger and how its event receiver is configured.

Secrets, identity, and supply chain

The guide covers storing and rotating secrets with Secret Manager and Cloud KMS, authenticating with JWT, OAuth 2.0, Identity Platform, or Workload Identity Federation, and protecting services with IAP. On the build side, you configure provenance in Cloud Build and enforce it with Binary Authorization, and respond to findings from Artifact Analysis and Security Command Center.

Data consistency and access from code

You need to understand eventual versus strong consistency across AlloyDB, Bigtable, Cloud SQL, Spanner, and Cloud Storage, and design schemas for structured and unstructured stores. Smaller items appear too, such as issuing signed URLs for Cloud Storage objects and writing data to BigQuery for analytics and AI/ML workloads.

Why CertSharp

Built for Exam Day, Not Just Study Time

Study guides teach concepts. We prepare you for the actual PCD exam.

1

Cloud Run and GKE deployment scenarios cover the platform tradeoffs, traffic splitting, probes, and autoscaling behavior the exam tests

2

Security questions go deep on Secret Manager, Workload Identity Federation, Application Default Credentials, and Binary Authorization

3

Every explanation says why each wrong option fails, so you learn the distinction instead of memorizing the answer

4

Mapped to the current exam guide, including Cloud Build, Artifact Registry, Eventarc, Pub/Sub, and Google Cloud Observability

Free Sample Questions

PCD Practice Questions (Free Sample)

Real PCD exam-style questions from our 500-question Google Professional Cloud Developer bank, each with the correct answer and a detailed explanation.

Question 1 — PCD

A Cloud Run service sends a response and then continues a background thread to flush metrics, but the flush often stalls. What is the cause and fix?

  • A

    CPU is throttled outside request handling; use instance-based billing so CPU is allocated outside requests

  • B

    Concurrency is too high for threads; set concurrency to 1 to keep the process active

  • C

    The container is restarted after each response; add a liveness probe to prevent restarts

  • D

    Memory is released after each response; raise the memory limit to the largest value offered

Correct answer: A — Explanation

With request-based billing, CPU is guaranteed while a request is being processed, so work after the response is throttled. Selecting CPU always allocated, which is instance-based billing, keeps CPU available in the background. Memory, concurrency, and probes do not change CPU throttling behavior.

  1. Question 2Designing highly scalable, secure, and reliable cloud-native applicationsEasy

    A company wants to ensure just images that passed CI tests and vulnerability checks are deployed to GKE. Which service enforces this?

    • A

      Artifact Analysis by itself, because it automatically blocks deployment of risky images

    • B

      Cloud Armor, by adding a rule on traffic to the cluster

    • C

      Binary Authorization with an attestor, requiring signed attestations from the pipeline

    • D

      IAM conditions on the cluster, which inspect image contents before deployment

    Show answer & explanation

    Correct answer: C — Binary Authorization with an attestor, requiring signed attestations from the pipeline

    Binary Authorization checks policy at deploy time, and attestations signed with a KMS-backed key prove the pipeline approved the image. Artifact Analysis reports vulnerabilities and does not block deployments by itself. Cloud Armor filters requests. IAM conditions cannot evaluate image contents.

  2. Question 3Integrating applications with Google Cloud servicesMedium

    A client receives a 400 INVALID_ARGUMENT response when creating a resource with a malformed name. What should the retry policy do?

    • A

      Not retry, log the error, and fix the request because the same payload fails identically each time

    • B

      Retry five times with exponential backoff in case the validation service is temporarily out of sync

    • C

      Retry immediately with the same payload and increase the timeout to give the server more time

    • D

      Switch to the gRPC transport and resend, since REST validation rules are stricter than gRPC rules

    Show answer & explanation

    Correct answer: A — Not retry, log the error, and fix the request because the same payload fails identically each time

    A 400 is a client error caused by a malformed request, so resending the same input yields the same result. Backoff helps with transient server or quota conditions. A longer timeout cannot fix invalid input. Both transports apply the same validation to the same API method.

  3. Question 4Configuring cloud-native applications for deploymentEasy

    A Deployment serves production traffic with 6 replicas. During a version update, the team must keep full serving capacity, so no old Pod may be removed before its replacement is Ready. Which rolling update settings achieve this?

    • A

      Set strategy type to Recreate so the new ReplicaSet starts first and the old ReplicaSet is removed afterwards

    • B

      Set maxUnavailable to 0 and maxSurge to 1 so a new Pod must become Ready before an old one is terminated

    • C

      Set maxUnavailable to 100 percent so the controller can replace the Pods in a single quick batch step

    • D

      Set minReadySeconds to 0 and delete the old Pods manually with kubectl delete pod once the image is pushed

    Show answer & explanation

    Correct answer: B — Set maxUnavailable to 0 and maxSurge to 1 so a new Pod must become Ready before an old one is terminated

    With maxUnavailable at 0 and maxSurge at 1, the controller creates an extra Pod and waits for readiness before it scales down an old Pod, so capacity never drops below 6. Recreate terminates the old Pods before starting new ones, causing downtime. A 100 percent maxUnavailable allows the whole set to go away at once. Manual deletion bypasses the rollout controller and gives no readiness gating.

  4. Question 5Building and testing applicationsEasy

    A cloudbuild.yaml has a unit test step followed by a docker build step. A unit test fails. What happens to the build?

    • A

      The build continues to the docker build step and marks the failure as a warning in the console

    • B

      The failing step returns a non-zero exit code, so the build fails and later steps do not run

    • C

      Cloud Build retries the failing step automatically until it passes or the project quota runs out

    • D

      The build skips the failed step and runs later steps in parallel to save time on the pipeline

    Show answer & explanation

    Correct answer: B — The failing step returns a non-zero exit code, so the build fails and later steps do not run

    A step that exits non-zero fails the whole build and subsequent steps are skipped, which makes test steps a natural quality gate before image creation. Cloud Build does not downgrade failures to warnings, retry steps on its own, or reorder the remaining steps.

  5. Question 6Designing highly scalable, secure, and reliable cloud-native applicationsMedium

    A Cloud Run service A calls a private Cloud Run service B in the same project. What is the correct authentication setup?

    • A

      Send an OAuth access token with the cloud-platform scope in the Authorization header

    • B

      Make B public and rely on the obscurity of its generated URL for protection

    • C

      Pass A's service account key as a basic auth header on each outbound request

    • D

      Grant A's service account Invoker on B and send a Google-signed ID token for B's URL

    Show answer & explanation

    Correct answer: D — Grant A's service account Invoker on B and send a Google-signed ID token for B's URL

    Private Cloud Run services accept Google-signed ID tokens whose audience is the target URL, from callers with the Invoker role. Access tokens are meant for Google APIs, not for authenticating to Cloud Run. A public service has no access control. Keys in headers are unsupported and weak.

  6. Question 7Integrating applications with Google Cloud servicesHard

    A game increments a global score counter stored in one Firestore document thousands of times per second, and writes begin to fail with contention errors. Which design resolves this?

    • A

      Add a composite index on the counter field so that updates are routed to a faster storage tier

    • B

      Split the counter into several shard documents, increment a random shard, and sum shards on read

    • C

      Wrap each increment in a transaction so that the document lock is held for a shorter period of time

    • D

      Move the counter into a field of the user profile documents so that the writes stay in one collection

    Show answer & explanation

    Correct answer: B — Split the counter into several shard documents, increment a random shard, and sum shards on read

    A single document supports only a limited sustained write rate, so a hot counter needs a distributed counter: many shard documents incremented at random and summed when read. Transactions add contention rather than removing it. Indexes do not raise per-document write throughput. Moving the field elsewhere does not change that it is a hot document.

  7. Question 8Configuring cloud-native applications for deploymentMedium

    Cloud Scheduler must start a Cloud Run job every night. How should the scheduler authenticate to the Cloud Run Admin API run endpoint?

    • A

      Use an HTTP target with an OIDC token whose audience is the name of the job being started

    • B

      Use a Pub/Sub target, because Cloud Scheduler cannot call Google APIs directly over HTTP targets

    • C

      Use an HTTP target with an OAuth token for a service account allowed to run the job, as it is a Google API

    • D

      Use an unauthenticated HTTP target, since the Admin API trusts requests from Cloud Scheduler

    Show answer & explanation

    Correct answer: C — Use an HTTP target with an OAuth token for a service account allowed to run the job, as it is a Google API

    Calls to googleapis.com endpoints expect OAuth access tokens, which Cloud Scheduler can attach for a chosen service account holding the permissions to run the job. OIDC tokens are meant for calling services such as Cloud Run endpoints, the API is not unauthenticated, and HTTP targets can call Google APIs.

  8. Question 9Building and testing applicationsMedium

    A monorepo holds several services. Changes under services/payments/ should trigger the payments build, and documentation edits should not trigger anything. How should the trigger be set up?

    • A

      Create a trigger with a branch regex of payments so a branch literally named payments starts builds

    • B

      Add an if statement to the first step that exits with code zero when the diff contains no payments files

    • C

      Set an included files filter of services/payments/** on the trigger, which skips builds when no matching file changed

    • D

      Disable the trigger and ask developers to run gcloud builds submit with a directory argument for payments

    Show answer & explanation

    Correct answer: C — Set an included files filter of services/payments/** on the trigger, which skips builds when no matching file changed

    Included files filters use glob patterns so the trigger fires only when a changed file matches, skipping unrelated commits at no build cost. A branch regex matches branch names, not paths. A step that exits successfully still consumes a build and reports success. Manual submits remove automation.

  9. Question 10Designing highly scalable, secure, and reliable cloud-native applicationsHard

    A workload identity pool trusts GitHub's OIDC provider, and a service account is bound so that every repository in the organization can impersonate it. How should the team restrict it to one repository?

    • A

      Restrict the audience of the OIDC provider to the name of the GitHub organization

    • B

      Add an attribute condition on the repository claim and bind a principalSet to that attribute

    • C

      Bind the Workload Identity User role to the individual GitHub username of each developer

    • D

      Delete the pool and configure the workflow to use a long-lived service account key instead

    Show answer & explanation

    Correct answer: B — Add an attribute condition on the repository claim and bind a principalSet to that attribute

    Attribute mapping plus an attribute condition on the repository claim, with a principalSet bound to that attribute, limits impersonation to the intended repo. The audience identifies the relying party and does not narrow the repo. Usernames are not the identity of a workflow. A key replaces federation with a weaker credential.

Want more PCD practice questions?

Take the free PCD practice test: 30 questions per session with instant explanations. Unlock all 500 questions plus timed exam mode for $11.99 one-time.

Start the Free 30-Question Practice Test

Pricing

Choose Your Plan

Start free, upgrade when you're ready

Free

Good for exploring the platform

$0
  • 30 practice questions
  • Practice mode only
  • Progress tracking
  • No exam simulation mode
Start Free
POPULAR

PCD Access

Best for the last few weeks before your PCD exam, once you have built and deployed a service yourself

$11.99one-time
  • 500 practice questions
  • Practice & Exam modes
  • Detailed explanations
  • Lifetime access
Get PCD Access — $11.99

7-day money-back guarantee

BEST VALUE

Pro — All Certs

Planning multiple certifications

$9.99/month
  • ALL certifications included
  • Unlimited questions
  • New certs added free
  • Cancel anytime
View Pro Plans

Save 30% vs buying individually

Your PCD Exam Won't Wait

Score 80% on two full CertSharp mock exams and you are calibrated to pass the live PCD exam.

Don't risk $200 on the real exam without testing yourself first.

Start Free Practice Now

No credit card required · 30 free questions · Upgrade for $11.99 when ready

Official Certification Resources

Our practice questions align with official exam objectives. Study official materials first, then use CertSharp to test your readiness.

View Official PCD Exam Details

FAQ

PCD Frequently Asked Questions

Everything you need to know about Google Professional Cloud Developer

What is the Google Professional Cloud Developer certification?
The Google Professional Cloud Developer (PCD) validates your ability to build and configure scalable, secure cloud-native applications on Google Cloud. It covers application design, building and testing, deployment to Cloud Run and GKE, and integration with Google Cloud services. Cost: $200, 50–60 questions, 2 hours.
How hard is the Google Professional Cloud Developer exam?
PCD is a professional-level exam and is harder than Associate Cloud Engineer. Questions are scenario-based and ask you to choose the best design, configuration, or fix for an application, so you need to know the tradeoffs between services such as Cloud Run, GKE, Pub/Sub, and Eventarc, plus secure authentication patterns. Hands-on development experience makes the biggest difference.
Is the Google Professional Cloud Developer worth it in 2026?
It is a good fit if you build or maintain applications on Google Cloud. It signals that you can design, test, deploy, and troubleshoot cloud-native applications, not just administer infrastructure. Its value depends on your role, so check whether the jobs you are targeting list Google Cloud development skills.
What experience do I need before taking Professional Cloud Developer?
Google lists no prerequisites. It recommends 3+ years of industry experience including 1+ years designing and managing solutions using Google Cloud. Experience writing and deploying containerized applications is especially useful.
What topics are on the Google Professional Cloud Developer exam?
The exam guide has four sections: designing highly scalable, secure, and reliable cloud-native applications (~32%), building and testing applications (~23%), configuring cloud-native applications for deployment (~24%), and integrating applications with Google Cloud services (~21%). Key topics include Cloud Run, GKE, Pub/Sub, Eventarc, Secret Manager, IAM and service accounts, Cloud Build, Artifact Registry, Cloud Storage, Firestore, Cloud SQL, and observability.
How many questions are on the Google Professional Cloud Developer exam?
Google states the exam has 50-60 multiple choice and multiple select questions and lasts two hours. Google does not publish a passing score, so aim for consistent, comfortable results across all four sections rather than a single target number.
How long to prepare for Google Professional Cloud Developer?
It depends on your background. Candidates who already build on Google Cloud often plan several weeks of focused review, while those newer to the platform need more time with hands-on labs. Spend the most time on the largest section, designing scalable, secure, and reliable applications (~32%), and on deployment to Cloud Run and GKE (~24%).
What is the difference between Professional Cloud Developer and Associate Cloud Engineer?
Associate Cloud Engineer (ACE) focuses on deploying and operating Google Cloud resources: projects, IAM, networking, and monitoring. Professional Cloud Developer focuses on building applications: writing and testing code, containerizing and deploying it to Cloud Run and GKE, and integrating it with Google Cloud APIs and services. They are independent exams, and ACE is not a prerequisite.
How much does the Google Professional Cloud Developer exam cost?
The Google Professional Cloud Developer exam costs $200 USD, plus tax where applicable. You can take it online-proctored from a remote location or at a testing center. It is offered in English and Japanese.
Is there a free Google Professional Cloud Developer practice test?
Yes. CertSharp offers a free Google Professional Cloud Developer practice test: 30 questions per session with detailed explanations, no credit card required. The full 500-question Google Professional Cloud Developer bank, including timed exam mode, is $11.99 one-time with lifetime access.