Pass Google Professional Cloud Architect Exam in First Attempt Easily
Real Google Professional Cloud Architect Exam Questions, Accurate & Verified Answers As Experienced in the Actual Test!

Verified by experts
3 products

You save $69.98

Professional Cloud Architect Premium Bundle

  • Premium File 370 Questions & Answers
  • Last Update: Nov 8, 2026
  • Training Course 63 Lectures
  • Study Guide 491 Pages
$79.99 $149.97

Purchase Individually

  • Premium File

    370 Questions & Answers
    Last Update: Nov 8, 2026

    $76.99
    $69.99
  • Training Course

    63 Lectures

    $43.99
    $39.99
  • Study Guide

    491 Pages

    $43.99
    $39.99

Google Cloud Certified - Professional Cloud Architect Professional Cloud Architect Practice Test 18

Question 170

A cloud platform team hosts central networking in a separate host project and needs application teams to administer their own compute but not VPC networks. Which mechanism best explains the failure mode in this situation?

  1. A service project automatically inherits authority to change host-project networking resources.
  2. Shared VPC requires all application workloads and firewall policies to reside in one project.
  3. Network User on a subnet is equivalent to full permission to modify the subnet's routes.
  4. Shared VPC separates network ownership from application resources, preventing each team from creating unrestricted independent routing domains.

Explanation

Correct Answer

D:

Shared VPC separates network ownership from application resources, preventing each team from creating unrestricted independent routing domains. That causal principle explains the original failure and supports the recommended intervention: Use Shared VPC with host/service project separation and narrowly granted network-user roles on approved subnets. An observed effect alone is not a substitute for the mechanism.

Incorrect Answers

A:

A service project automatically inherits authority to change host-project networking resources. It offers an appealing simple explanation but attributes the problem to a condition not established by the source-case facts. The mechanism that actually connects the requirement to the architecture is: Shared VPC separates network ownership from application resources, preventing each team from creating unrestricted independent routing domains.

B:

Shared VPC requires all application workloads and firewall policies to reside in one project. This overstates an architectural component or confuses a convenient symptom with its cause. The case instead depends on: shared VPC separates network ownership from application resources, preventing each team from creating unrestricted independent routing domains. This is why the appropriate intervention is: Use Shared VPC with host/service project separation and narrowly granted network-user roles on approved subnets.

C:

Network User on a subnet is equivalent to full permission to modify the subnet's routes. The claim would leave the essential causal condition unexplained for this scenario: A cloud platform team hosts central networking in a separate host project and needs application teams to administer their own compute but not VPC networks. The correct underlying technical relationship is: Shared VPC separates network ownership from application resources, preventing each team from creating unrestricted independent routing domains. Source-case challenge: Network User permission to attach to a subnet does not itself imply full ability to modify routes or firewall rules.

Question 171

A financial API sees poor tail latency although mean latency is acceptable and compute utilization is moderate. What remaining exposure should the architecture review continue to track?

  1. A new dependency or traffic mix can recreate tail latency after the measured optimization.
  2. A future API version might change the format of collected trace attributes.
  3. An observability sampling rule could obscure an infrequent but unrelated 400-class error.
  4. The incident response dashboard might not have the access controls required for sensitive trace metadata.

Explanation

Correct Answer

A:

A new dependency or traffic mix can recreate tail latency after the measured optimization. Even when the immediate design intervention succeeds, a later condition can still invalidate its assumptions. This scenario requires monitoring that specific dependency rather than a generic project concern.

Incorrect Answers

B:

A future API version might change the format of collected trace attributes. That is a conceivable secondary concern, but the most direct residual after the proposed change is tied to the remaining unaddressed dependency: A new dependency or traffic mix can recreate tail latency after the measured optimization. The scenario is specifically: A financial API sees poor tail latency although mean latency is acceptable and compute utilization is moderate.

C:

An observability sampling rule could obscure an infrequent but unrelated 400-class error. It deserves normal operating controls, yet it is less directly evidenced by this case than: A new dependency or traffic mix can recreate tail latency after the measured optimization. The relevant technical boundary is: Averages hide a long tail caused by specific downstream waits, so changing VM size before finding the bottleneck wastes capacity. Source-case challenge: Log wording changes are not a demonstrated risk; changed dependency timing or traffic mix can recreate user-visible tail failures.

D:

The incident response dashboard might not have the access controls required for sensitive trace metadata. This could matter in a different release or configuration, but it is not the strongest case-supported unresolved exposure. A new dependency or traffic mix can recreate tail latency after the measured optimization is more specifically connected to the architecture because averages hide a long tail caused by specific downstream waits, so changing VM size before finding the bottleneck wastes capacity.

Question 172

A software supply-chain pipeline publishes unsigned containers to Artifact Registry and the security team cannot determine the provenance of a deployed image. What design property is most important to understand before changing the implementation?

  1. A verified artifact signature or attestation binds deployed bytes to a trusted build process; mutable tags alone provide no trustworthy provenance.
  2. An image tag identifies exact immutable bytes even after a rebuild pushes the same tag.
  3. An unsigned artifact registry entry proves trusted supply-chain origin by existing in the registry.
  4. A security scan alone establishes that deployed bytes came from the approved CI pipeline.

Explanation

Correct Answer

A:

A verified artifact signature or attestation binds deployed bytes to a trusted build process; mutable tags alone provide no trustworthy provenance. That causal principle explains the original failure and supports the recommended intervention: Use trusted build identities, record provenance and attestations, and enforce policy-based image admission before GKE deployment. An observed effect alone is not a substitute for the mechanism.

Incorrect Answers

B:

An image tag identifies exact immutable bytes even after a rebuild pushes the same tag. It offers an appealing simple explanation but attributes the problem to a condition not established by the source-case facts. The mechanism that actually connects the requirement to the architecture is: A verified artifact signature or attestation binds deployed bytes to a trusted build process; mutable tags alone provide no trustworthy provenance. Source-case challenge: Image tags are pointers that may be overwritten; signed digests or attestations must bind deployed bytes to provenance.

C:

An unsigned artifact registry entry proves trusted supply-chain origin by existing in the registry. This overstates an architectural component or confuses a convenient symptom with its cause. The case instead depends on: a verified artifact signature or attestation binds deployed bytes to a trusted build process; mutable tags alone provide no trustworthy provenance. This is why the appropriate intervention is: Use trusted build identities, record provenance and attestations, and enforce policy-based image admission before GKE deployment.

D:

A security scan alone establishes that deployed bytes came from the approved CI pipeline. The claim would leave the essential causal condition unexplained for this scenario: A software supply-chain pipeline publishes unsigned containers to Artifact Registry and the security team cannot determine the provenance of a deployed image. The correct underlying technical relationship is: A verified artifact signature or attestation binds deployed bytes to a trusted build process; mutable tags alone provide no trustworthy provenance.

Question 173

A hybrid application resolves private database DNS inconsistently because developers configured public records and local static host files. What is the most direct acceptance evidence for this design decision?

  1. A workstation resolves the database correctly while on the office LAN.
  2. A public resolver returns an address but no private-path test is performed.
  3. The private zone exists in one VPC but the on-premises resolver query is never checked.
  4. Resolvers on both sides of the hybrid link return the approved internal address while public clients cannot resolve sensitive private records.

Explanation

Correct Answer

D:

The acceptance evidence is: Resolvers on both sides of the hybrid link return the approved internal address while public clients cannot resolve sensitive private records. This validates the important boundary or service outcome rather than relying on a design assertion. The technical reason for requiring this evidence is: Name resolution must follow explicit private-view routing; static client files can become stale and leak private names to public resolvers.

Incorrect Answers

A:

A workstation resolves the database correctly while on the office LAN. That observation can be collected, but it does not exercise the service-level or security outcome named in the requirement. More direct end-to-end acceptance evidence is: Resolvers on both sides of the hybrid link return the approved internal address while public clients cannot resolve sensitive private records. Source-case challenge: Local workstation resolution is only one vantage point; both sides must resolve the approved internal address without public leakage.

B:

A public resolver returns an address but no private-path test is performed. This is a limited configuration or proxy signal and could be green while the requested outcome still fails. A representative successful test instead requires: Resolvers on both sides of the hybrid link return the approved internal address while public clients cannot resolve sensitive private records. The causal issue is: Name resolution must follow explicit private-view routing; static client files can become stale and leak private names to public resolvers.

C:

The private zone exists in one VPC but the on-premises resolver query is never checked. The outcome being observed is not equivalent to the original user requirement for a transaction, control or recovery property. The specific proof must show: Resolvers on both sides of the hybrid link return the approved internal address while public clients cannot resolve sensitive private records.

Question 174

A developer writes custom retries for Google APIs but retries every HTTP error indefinitely, creating duplicate billing operations. Which test provides the strongest evidence that the proposed remedy succeeded?

  1. The API call eventually returns 200 after a long sequence of retries.
  2. A retry test counts requests but never checks duplicate charge records.
  3. Injected 429 and 503 responses trigger bounded retries, while invalid credentials fail fast and a repeated request produces no duplicate business action.
  4. A one-off successful call proves credentials work while authorization errors loop forever.

Explanation

Correct Answer

C:

The acceptance evidence is: Injected 429 and 503 responses trigger bounded retries, while invalid credentials fail fast and a repeated request produces no duplicate business action. This validates the important boundary or service outcome rather than relying on a design assertion. The technical reason for requiring this evidence is: Transient retryable errors differ from permanent authorization errors, and nonidempotent mutation replay can duplicate effects.

Incorrect Answers

A:

The API call eventually returns 200 after a long sequence of retries. That observation can be collected, but it does not exercise the service-level or security outcome named in the requirement. More direct end-to-end acceptance evidence is: Injected 429 and 503 responses trigger bounded retries, while invalid credentials fail fast and a repeated request produces no duplicate business action.

B:

A retry test counts requests but never checks duplicate charge records. This is a limited configuration or proxy signal and could be green while the requested outcome still fails. A representative successful test instead requires: Injected 429 and 503 responses trigger bounded retries, while invalid credentials fail fast and a repeated request produces no duplicate business action. The causal issue is: Transient retryable errors differ from permanent authorization errors, and nonidempotent mutation replay can duplicate effects. Source-case challenge: Counting retries without reconciling charges does not establish one logical billing operation per client request.

D:

A one-off successful call proves credentials work while authorization errors loop forever. The outcome being observed is not equivalent to the original user requirement for a transaction, control or recovery property. The specific proof must show: Injected 429 and 503 responses trigger bounded retries, while invalid credentials fail fast and a repeated request produces no duplicate business action.

Question 175

A research group will train a large multimodal model with GPUs, but previous jobs spent half their duration waiting for data from a distant storage region. Which architecture recommendation best addresses the stated requirements and constraints?

  1. Co-locate appropriate accelerators and high-throughput data inputs, benchmark accelerator utilization, and adjust distributed training topology.
  2. Double the GPU count without addressing slow reads from the distant storage region.
  3. Move dataset preprocessing to the same distant region as before and measure only peak FLOPS.
  4. Increase memory allocation per GPU without profiling stalled input pipeline stages.

Explanation

Correct Answer

A:

The architecture should co-locate appropriate accelerators and high-throughput data inputs, benchmark accelerator utilization, and adjust distributed training topology. This addresses the scenario directly because data pipeline throughput and placement can dominate expensive accelerator utilization; more GPUs alone do not fix stalled input queues. The implementation is acceptable only if the team demonstrates: A representative training epoch shows higher device utilization and lower dollars per effective training step at acceptable model quality.

Incorrect Answers

B:

Double the GPU count without addressing slow reads from the distant storage region. This option changes a visible part of the architecture, but its scope is insufficient for the actual constraints: A research group will train a large multimodal model with GPUs, but previous jobs spent half their duration waiting for data from a distant storage region. The decisive dependency is that data pipeline throughput and placement can dominate expensive accelerator utilization; more GPUs alone do not fix stalled input queues. The better complete recommendation is: Co-locate appropriate accelerators and high-throughput data inputs, benchmark accelerator utilization, and adjust distributed training topology. Source-case challenge: More GPUs cannot fix a distant dataset bottleneck when current expensive devices already wait half the run for input.

C:

Move dataset preprocessing to the same distant region as before and measure only peak FLOPS. The proposed shortcut does not preserve every stated business or technical boundary in this case. In particular, data pipeline throughput and placement can dominate expensive accelerator utilization; more GPUs alone do not fix stalled input queues. The defensible decision is: Co-locate appropriate accelerators and high-throughput data inputs, benchmark accelerator utilization, and adjust distributed training topology. An implementation must be judged against: a representative training epoch shows higher device utilization and lower dollars per effective training step at acceptable model quality.

D:

Increase memory allocation per GPU without profiling stalled input pipeline stages. This would focus the effort on a different subsystem or presumed safeguard rather than the documented architecture need: A research group will train a large multimodal model with GPUs, but previous jobs spent half their duration waiting for data from a distant storage region. Co-locate appropriate accelerators and high-throughput data inputs, benchmark accelerator utilization, and adjust distributed training topology. is stronger because data pipeline throughput and placement can dominate expensive accelerator utilization; more GPUs alone do not fix stalled input queues.

Question 176

A deployment pipeline outside Google Cloud uses a downloadable long-lived service account key to write artifacts, and a developer accidentally commits the key to a public repository. Which solution balances the business and technical requirements most defensibly?

  1. Store a replacement long-lived JSON key in a private repository and rotate it only annually.
  2. Grant all external CI jobs Service Account Token Creator on every project.
  3. Trust any token issued by the external CI provider without restrictive repository and branch attributes.
  4. Remove and revoke exposed keys, configure Workload Identity Federation for the external CI identity, and constrain impersonated service-account permissions.

Explanation

Correct Answer

D:

The architecture should remove and revoke exposed keys, configure Workload Identity Federation for the external CI identity, and constrain impersonated service-account permissions. This addresses the scenario directly because federated short-lived credentials reduce reusable secret exposure while principal and attribute restrictions define which external workload is trusted. The implementation is acceptable only if the team demonstrates: An allowed signed CI identity deploys with temporary credentials, but an untrusted repository token is denied impersonation and no key remains stored.

Incorrect Answers

A:

Store a replacement long-lived JSON key in a private repository and rotate it only annually. This option changes a visible part of the architecture, but its scope is insufficient for the actual constraints: A deployment pipeline outside Google Cloud uses a downloadable long-lived service account key to write artifacts, and a developer accidentally commits the key to a public repository. The decisive dependency is that federated short-lived credentials reduce reusable secret exposure while principal and attribute restrictions define which external workload is trusted. The better complete recommendation is: Remove and revoke exposed keys, configure Workload Identity Federation for the external CI identity, and constrain impersonated service-account permissions.

B:

Grant all external CI jobs Service Account Token Creator on every project. The proposed shortcut does not preserve every stated business or technical boundary in this case. In particular, federated short-lived credentials reduce reusable secret exposure while principal and attribute restrictions define which external workload is trusted. The defensible decision is: Remove and revoke exposed keys, configure Workload Identity Federation for the external CI identity, and constrain impersonated service-account permissions. An implementation must be judged against: an allowed signed CI identity deploys with temporary credentials, but an untrusted repository token is denied impersonation and no key remains stored. Source-case challenge: Granting every external CI workflow broad impersonation defeats the intention of source-specific short-lived trust.

C:

Trust any token issued by the external CI provider without restrictive repository and branch attributes. This would focus the effort on a different subsystem or presumed safeguard rather than the documented architecture need: A deployment pipeline outside Google Cloud uses a downloadable long-lived service account key to write artifacts, and a developer accidentally commits the key to a public repository. Remove and revoke exposed keys, configure Workload Identity Federation for the external CI identity, and constrain impersonated service-account permissions. is stronger because federated short-lived credentials reduce reusable secret exposure while principal and attribute restrictions define which external workload is trusted.

Question 177

A development group tests Firestore event logic against production because it has no practical local integration environment. Which design choice most directly closes the demonstrated gap?

  1. Run Firestore event tests against production with a test-only document prefix.
  2. Assume the local emulator reproduces cloud IAM and quota behavior perfectly.
  3. Use a supported local Firestore emulator in CI with deterministic fixtures and supplement it with controlled deployed integration tests.
  4. Test only Firestore API object creation with no event ordering or retry cases.

Explanation

Correct Answer

C:

The architecture should use a supported local Firestore emulator in CI with deterministic fixtures and supplement it with controlled deployed integration tests. This addresses the scenario directly because an emulator speeds reproducible tests without production data changes but cannot perfectly reproduce cloud IAM, quotas or all managed-service behavior. The implementation is acceptable only if the team demonstrates: Automated tests pass on the emulator and a separate authorized staging test confirms real identity and network behavior.

Incorrect Answers

A:

Run Firestore event tests against production with a test-only document prefix. This option changes a visible part of the architecture, but its scope is insufficient for the actual constraints: A development group tests Firestore event logic against production because it has no practical local integration environment. The decisive dependency is that an emulator speeds reproducible tests without production data changes but cannot perfectly reproduce cloud IAM, quotas or all managed-service behavior. The better complete recommendation is: Use a supported local Firestore emulator in CI with deterministic fixtures and supplement it with controlled deployed integration tests. Source-case challenge: Testing events against production—even with prefixed documents—can trigger real side effects or expose user data.

B:

Assume the local emulator reproduces cloud IAM and quota behavior perfectly. The proposed shortcut does not preserve every stated business or technical boundary in this case. In particular, an emulator speeds reproducible tests without production data changes but cannot perfectly reproduce cloud IAM, quotas or all managed-service behavior. The defensible decision is: Use a supported local Firestore emulator in CI with deterministic fixtures and supplement it with controlled deployed integration tests. An implementation must be judged against: automated tests pass on the emulator and a separate authorized staging test confirms real identity and network behavior.

D:

Test only Firestore API object creation with no event ordering or retry cases. This would focus the effort on a different subsystem or presumed safeguard rather than the documented architecture need: A development group tests Firestore event logic against production because it has no practical local integration environment. Use a supported local Firestore emulator in CI with deterministic fixtures and supplement it with controlled deployed integration tests. is stronger because an emulator speeds reproducible tests without production data changes but cannot perfectly reproduce cloud IAM, quotas or all managed-service behavior.

Question 178

A company completes monthly penetration tests but its team asserts that those tests demonstrate its new capacity autoscaler can handle promotional traffic. How should the team demonstrate the intended security or reliability outcome?

  1. The scanner reports no high-severity vulnerabilities in the new release.
  2. A representative traffic spike meets SLO and controlled adversarial probes cannot access protected paths.
  3. A lone client request succeeds without a production-scale workload replay.
  4. All pods pass readiness probes before the sale but target SLOs are not measured under load.

Explanation

Correct Answer

B:

The acceptance evidence is: A representative traffic spike meets SLO and controlled adversarial probes cannot access protected paths. This validates the important boundary or service outcome rather than relying on a design assertion. The technical reason for requiring this evidence is: Security attack simulations and load-capacity exercises test different properties; a passing vulnerability scan is not throughput evidence.

Incorrect Answers

A:

The scanner reports no high-severity vulnerabilities in the new release. That observation can be collected, but it does not exercise the service-level or security outcome named in the requirement. More direct end-to-end acceptance evidence is: A representative traffic spike meets SLO and controlled adversarial probes cannot access protected paths.

C:

A lone client request succeeds without a production-scale workload replay. This is a limited configuration or proxy signal and could be green while the requested outcome still fails. A representative successful test instead requires: A representative traffic spike meets SLO and controlled adversarial probes cannot access protected paths. The causal issue is: Security attack simulations and load-capacity exercises test different properties; a passing vulnerability scan is not throughput evidence.

D:

All pods pass readiness probes before the sale but target SLOs are not measured under load. The outcome being observed is not equivalent to the original user requirement for a transaction, control or recovery property. The specific proof must show: A representative traffic spike meets SLO and controlled adversarial probes cannot access protected paths. Source-case challenge: Healthy readiness probes are insufficient if performance under the promotional traffic level has never been measured.

Question 179

A globally deployed API claims resilience because it has several zones but its engineers have never tested losing a zone under live-like load. Which mechanism best explains the failure mode in this situation?

  1. Provisioning nodes in several zones proves real customer requests survive a zonal outage.
  2. An architecture diagram establishes failover timing without load and state checks.
  3. A successful synthetic ping is equivalent to an authenticated transaction during failover.
  4. Redundant components do not demonstrate effective failover until application routing and dependencies survive a representative fault.

Explanation

Correct Answer

D:

Redundant components do not demonstrate effective failover until application routing and dependencies survive a representative fault. That causal principle explains the original failure and supports the recommended intervention: Run approved fault-injection exercises against isolated or safeguarded production traffic with stop conditions, and evaluate service-level indicators. An observed effect alone is not a substitute for the mechanism.

Incorrect Answers

A:

Provisioning nodes in several zones proves real customer requests survive a zonal outage. It offers an appealing simple explanation but attributes the problem to a condition not established by the source-case facts. The mechanism that actually connects the requirement to the architecture is: Redundant components do not demonstrate effective failover until application routing and dependencies survive a representative fault. Source-case challenge: Having multiple zones is architectural redundancy, not proof that the complete application meets an SLO during a fault.

B:

An architecture diagram establishes failover timing without load and state checks. This overstates an architectural component or confuses a convenient symptom with its cause. The case instead depends on: redundant components do not demonstrate effective failover until application routing and dependencies survive a representative fault. This is why the appropriate intervention is: Run approved fault-injection exercises against isolated or safeguarded production traffic with stop conditions, and evaluate service-level indicators.

C:

A successful synthetic ping is equivalent to an authenticated transaction during failover. The claim would leave the essential causal condition unexplained for this scenario: A globally deployed API claims resilience because it has several zones but its engineers have never tested losing a zone under live-like load. The correct underlying technical relationship is: Redundant components do not demonstrate effective failover until application routing and dependencies survive a representative fault.

Question 180

A national pharmacy retailer expects a six-hour promotional traffic spike but must retain its existing payment processor and avoid idle servers for the rest of the month. After the recommended changes, which residual risk is most directly relevant?

  1. The load balancer's TLS renewal could require a separately rehearsed change window.
  2. Payment fraud-screening rules might reject unusually large promotional orders.
  3. An untested identity-provider dependency might affect new checkout sessions during a region failure.
  4. The external processor can remain a throughput bottleneck even when the storefront scales correctly.

Explanation

Correct Answer

D:

The external processor can remain a throughput bottleneck even when the storefront scales correctly. Even when the immediate design intervention succeeds, a later condition can still invalidate its assumptions. This scenario requires monitoring that specific dependency rather than a generic project concern.

Incorrect Answers

A:

The load balancer's TLS renewal could require a separately rehearsed change window. That is a conceivable secondary concern, but the most direct residual after the proposed change is tied to the remaining unaddressed dependency: The external processor can remain a throughput bottleneck even when the storefront scales correctly. The scenario is specifically: A national pharmacy retailer expects a six-hour promotional traffic spike but must retain its existing payment processor and avoid idle servers for the rest of the month.

B:

Payment fraud-screening rules might reject unusually large promotional orders. It deserves normal operating controls, yet it is less directly evidenced by this case than: The external processor can remain a throughput bottleneck even when the storefront scales correctly. The relevant technical boundary is: Independently scaling the stateless web tier absorbs variable demand without replacing the regulated payment interface or paying for peak capacity continuously. Source-case challenge: Fraud-screening declines are possible, but the stipulated unchanged external processor throughput is the explicit architectural coupling left unresolved.

C:

An untested identity-provider dependency might affect new checkout sessions during a region failure. This could matter in a different release or configuration, but it is not the strongest case-supported unresolved exposure. The external processor can remain a throughput bottleneck even when the storefront scales correctly is more specifically connected to the architecture because independently scaling the stateless web tier absorbs variable demand without replacing the regulated payment interface or paying for peak capacity continuously.

Google Cloud Certified - Professional Cloud Architect (Professional Cloud Architect) Practice Tests

Purchase Individually

  • Premium File

    370 Questions & Answers
    Last Update: Nov 8, 2026

    $76.99
    $69.99
  • Training Course

    63 Lectures

    $43.99
    $39.99
  • Study Guide

    491 Pages

    $43.99
    $39.99

SPECIAL OFFER: GET 10% OFF
This is ONE TIME OFFER

You save
10%

Enter Your Email Address to Receive Your 10% Off Discount Code

SPECIAL OFFER: GET 10% OFF

You save
10%

Use Discount Code:

A confirmation link was sent to your e-mail.

Please check your mailbox for a message from support@examlabs.com and follow the directions.

Download Free Demo of VCE Exam Simulator

Experience Avanset VCE Exam Simulator for yourself.

Simply submit your email address below to get started with our interactive software demo of your free trial.

  • Realistic exam simulation and exam editor with preview functions
  • Whole exam in a single file with several different question types
  • Customizable exam-taking mode & detailed score reports