Google Professional Cloud Architect Practice Test 87
Question 860
Urban public transport authority plans to update fare validation backend. Investigators found that histogram shows 8x peak events and lock waits on identical card records, consistent with the reported problem that rush-hour taps overwhelm the settlement database even though fraud checks are light. The response must satisfy the requirement that a transit pass must not be charged twice for one tap. Select TWO complementary actions: one durable target architecture and one controlled way to introduce it while preserving service. Which TWO are best?
- Monitor pending tap age, settlement retries and contention
- Measure tap interarrival, hot-key contention and duplicate IDs
- Scale only the front-end web server while keeping the hot database lock
- Introduce tap ingestion in station cohorts with controlled retry semantics
- Buffer taps durably and enforce idempotent settlement with per-pass consistency
Explanation
Correct Answers
D:
Introduce tap ingestion in station cohorts with controlled retry semantics is a controlled means to introduce the approved architecture while meeting the requirement that a transit pass must not be charged twice for one tap. Staged delivery, attributable change evidence and rollback guard against an otherwise avoidable service disruption. This supplies the controlled-introduction half of the two-action package.
E:
Buffer taps durably and enforce idempotent settlement with per-pass consistency targets the underlying issue (rush-hour taps overwhelm the settlement database even though fraud checks are light) rather than merely reporting symptoms. It also preserves the required outcome (a transit pass must not be charged twice for one tap), which a target-state design must satisfy. This supplies the lasting-control half of the two-action package.
Incorrect Answers
A:
Monitoring fare validation backend for future drift is useful after the rollout, but it does not specify the requested target control and safe implementation for this case: rush-hour taps overwhelm the settlement database even though fraud checks are light.
B:
This diagnoses the original issue (rush-hour taps overwhelm the settlement database even though fraud checks are light), but the TWO requested actions are the target architecture and its implementation; discovery was already completed.
C:
A larger front end does not remove the transactional bottleneck. The shortcut does not supply the requested design and safe rollout for fare validation backend, where histogram shows 8x peak events and lock waits on identical card records.
Question 861
Telecom customer portal plans to update Google Cloud SDK workload identity. Investigators found that commit scan reveals key reuse across three build agents, consistent with the reported problem that a deployment script has a static JSON key checked into a Git repository. The response must satisfy the requirement that deployments need constrained short-lived credentials. Select TWO complementary actions: one durable target architecture and one controlled way to introduce it while preserving service. Which TWO are best?
- Monitor federation changes and unexpected credential usage
- Transition build jobs to federation and revoke leaked keys
- Copy the static key into the CI variable store without revoking it
- Use Workload Identity Federation with constrained provider attributes
- Inventory CI identity provider claims and project permissions
Explanation
Correct Answers
B:
Transition build jobs to federation and revoke leaked keys is a controlled means to introduce the approved architecture while meeting the requirement that deployments need constrained short-lived credentials. Staged delivery, attributable change evidence and rollback guard against an otherwise avoidable service disruption. This supplies the controlled-introduction half of the two-action package.
D:
Use Workload Identity Federation with constrained provider attributes targets the underlying issue (a deployment script has a static JSON key checked into a Git repository) rather than merely reporting symptoms. It also preserves the required outcome (deployments need constrained short-lived credentials), which a target-state design must satisfy. This supplies the lasting-control half of the two-action package.
Incorrect Answers
A:
Monitoring Google Cloud SDK workload identity for future drift is useful after the rollout, but it does not specify the requested target control and safe implementation for this case: a deployment script has a static JSON key checked into a Git repository.
C:
Relocating the leaked credential does not eliminate its misuse risk. The shortcut does not supply the requested design and safe rollout for Google Cloud SDK workload identity, where commit scan reveals key reuse across three build agents.
E:
This diagnoses the original issue (a deployment script has a static JSON key checked into a Git repository), but the TWO requested actions are the target architecture and its implementation; discovery was already completed.
Question 862
Case study: Public broadcaster: live event thumbnail pipeline
Public broadcaster relies on live event thumbnail pipeline to deliver its operational commitments. The current architectural problem is that image conversion jobs scale unpredictably and operators provision fixed VMs manually. The team has recorded this evidence: daily charts show 60x spikes lasting ten minutes with idle nights. Any proposed response must preserve the requirement that uploads require isolated execution and no long-lived server state.
Public broadcaster plans to update live event thumbnail pipeline. Investigators found that daily charts show 60x spikes lasting ten minutes with idle nights, consistent with the reported problem that image conversion jobs scale unpredictably and operators provision fixed VMs manually. The response must satisfy the requirement that uploads require isolated execution and no long-lived server state. Select TWO complementary actions: one durable target architecture and one controlled way to introduce it while preserving service. Which TWO are best?
- Classify request duration, concurrency, output size and idle baseline
- Monitor instance scaling, errors and cost per generated image
- Operate a fixed single VM with no scaling or restart policy
- Publish container revisions with gradual traffic splitting and bounded concurrency
- Select Cloud Run for stateless event-driven conversion with managed scaling
Explanation
Correct Answers
D:
Publish container revisions with gradual traffic splitting and bounded concurrency is a controlled means to introduce the approved architecture while meeting the requirement that uploads require isolated execution and no long-lived server state. Staged delivery, attributable change evidence and rollback guard against an otherwise avoidable service disruption. This supplies the controlled-introduction half of the two-action package.
E:
Select Cloud Run for stateless event-driven conversion with managed scaling targets the underlying issue (image conversion jobs scale unpredictably and operators provision fixed VMs manually) rather than merely reporting symptoms. It also preserves the required outcome (uploads require isolated execution and no long-lived server state), which a target-state design must satisfy. This supplies the lasting-control half of the two-action package.
Incorrect Answers
A:
This diagnoses the original issue (image conversion jobs scale unpredictably and operators provision fixed VMs manually), but the TWO requested actions are the target architecture and its implementation; discovery was already completed.
B:
Monitoring live event thumbnail pipeline for future drift is useful after the rollout, but it does not specify the requested target control and safe implementation for this case: image conversion jobs scale unpredictably and operators provision fixed VMs manually.
C:
A fixed VM creates capacity bottlenecks and idle cost. The shortcut does not supply the requested design and safe rollout for live event thumbnail pipeline, where daily charts show 60x spikes lasting ten minutes with idle nights.
Question 863
Case study: National airport lounge operator: managed Prometheus scraping
National airport lounge operator relies on managed Prometheus scraping to deliver its operational commitments. The current architectural problem is that several GKE services expose metrics but the monitoring system has blind spots after namespace moves. The team has recorded this evidence: scrape health shows targets disappeared when Kubernetes labels changed. Any proposed response must preserve the requirement that capacity alerts must remain tied to actual service discovery.
Consider National airport lounge operator and its managed Prometheus scraping workload. The issue is that several GKE services expose metrics but the monitoring system has blind spots after namespace moves; the current observation is that scrape health shows targets disappeared when Kubernetes labels changed. The team must ensure that capacity alerts must remain tied to actual service discovery. Which continuous monitoring approach best reveals recurrence or changing dependencies?
- Create a temporary test pod and verify metrics are collected and attributed
- Treat all unlabeled zero metrics as healthy application inactivity
- Watch missing targets, cardinality growth and failed scrapes
- Configure Managed Service for Prometheus collectors with stable target discovery
Explanation
Correct Answer
C:
Watch missing targets, cardinality growth and failed scrapes reveals recurrence, drift or changing dependencies after launch. It provides a repeatable signal for the risk (several GKE services expose metrics but the monitoring system has blind spots after namespace moves) rather than treating the first successful test as permanent assurance.
Incorrect Answers
A:
The alternative “Create a temporary test pod and verify metrics are collected and attributed” is a one-time acceptance result rather than ongoing operational assurance. In the reported observation, scrape health shows targets disappeared when Kubernetes labels changed. The option “Watch missing targets, cardinality growth and failed scrapes” instead addresses the requested ongoing operational assurance.
B:
Missing scrape targets can look like falsely healthy zeros. For this question, it is not a defensible choice for ongoing operational assurance and it does not satisfy the requirement that capacity alerts must remain tied to actual service discovery.
D:
The alternative “Configure Managed Service for Prometheus collectors with stable target discovery” selects a design, but does not provide the ongoing signal required to detect drift. In the reported observation, scrape health shows targets disappeared when Kubernetes labels changed. The option “Watch missing targets, cardinality growth and failed scrapes” instead addresses the requested ongoing operational assurance.
Question 864
At Global fleet analytics company, the connected vehicle insight platform workload has a documented risk: management expects ten times more telemetry after a merger but the platform allocates fixed partitions. The review found that forecast shows devices and event attributes will expand across regions. The architecture must also ensure that existing fleet alerts cannot lose events during schema evolution. What follow-up metric or alert will identify emerging failures after the initial acceptance test?
- Design partitioned event contracts and independently scalable analytical consumers
- Replay old and new telemetry versions through the same dashboards
- Watch partition hotspots, obsolete schemas and capacity forecasts
- Change all event schemas simultaneously without versioning
Explanation
Correct Answer
C:
Watch partition hotspots, obsolete schemas and capacity forecasts reveals recurrence, drift or changing dependencies after launch. It provides a repeatable signal for the risk (management expects ten times more telemetry after a merger but the platform allocates fixed partitions) rather than treating the first successful test as permanent assurance.
Incorrect Answers
A:
The alternative “Design partitioned event contracts and independently scalable analytical consumers” selects a design, but does not provide the ongoing signal required to detect drift. In the reported observation, forecast shows devices and event attributes will expand across regions. The option “Watch partition hotspots, obsolete schemas and capacity forecasts” instead addresses the requested ongoing operational assurance.
B:
The alternative “Replay old and new telemetry versions through the same dashboards” is a one-time acceptance result rather than ongoing operational assurance. In the reported observation, forecast shows devices and event attributes will expand across regions. The option “Watch partition hotspots, obsolete schemas and capacity forecasts” instead addresses the requested ongoing operational assurance.
D:
A schema cutover without compatibility can strand existing devices. For this question, it is not a defensible choice for ongoing operational assurance and it does not satisfy the requirement that existing fleet alerts cannot lose events during schema evolution.
Question 865
At Emergency veterinary service, the Google Cloud design for 24-hour booking network is affected because the migration estimate ignores the impact of a failed regional booking queue. A review found that incident records show bookings lost when the only queue broker restarts. The proposed response must account for the requirement that emergency hospitals must reconcile bookings after a site outage. Select TWO complementary actions: one durable target architecture and one controlled way to introduce it while preserving service. Which TWO are best?
- Build a durable booking pipeline with idempotent consumers and multi-zone service recovery
- Use an in-memory single-instance queue for all accepted bookings
- Calculate booking RPO/RTO, delivery guarantees and regional dependency chains
- Dual-write for a controlled pilot with deterministic booking identifiers
- Track queue depth, duplicate suppression and recovery timing
Explanation
Correct Answers
A:
Build a durable booking pipeline with idempotent consumers and multi-zone service recovery targets the underlying issue (the migration estimate ignores the impact of a failed regional booking queue) rather than merely reporting symptoms. It also preserves the required outcome (emergency hospitals must reconcile bookings after a site outage), which a target-state design must satisfy. This supplies the lasting-control half of the two-action package.
D:
Dual-write for a controlled pilot with deterministic booking identifiers is a controlled means to introduce the approved architecture while meeting the requirement that emergency hospitals must reconcile bookings after a site outage. Staged delivery, attributable change evidence and rollback guard against an otherwise avoidable service disruption. This supplies the controlled-introduction half of the two-action package.
Incorrect Answers
B:
A volatile single broker loses accepted work on restart. The shortcut does not supply the requested design and safe rollout for 24-hour booking network, where incident records show bookings lost when the only queue broker restarts.
C:
This diagnoses the original issue (the migration estimate ignores the impact of a failed regional booking queue), but the TWO requested actions are the target architecture and its implementation; discovery was already completed.
E:
Monitoring 24-hour booking network for future drift is useful after the rollout, but it does not specify the requested target control and safe implementation for this case: the migration estimate ignores the impact of a failed regional booking queue.
Question 866
At Public grant agency, the Google Cloud design for executive decision analysis is affected because project review scores optional chatbot convenience higher than accessibility and regulatory filing. A review found that business case omits users reliant on assistive technologies. The proposed response must account for the requirement that benefit delivery must meet accessible filing requirements. Which analysis will allow the architect to distinguish the reported symptom from the actual failure path?
- Prioritize accessible essential workflow before optional AI enhancements
- Translate user needs into measurable outcomes, compliance and risk
- Sequence the accessibility remediation with user testing and business ownership
- Approve the chatbot solely because it uses the newest language model
Explanation
Correct Answer
B:
Translate user needs into measurable outcomes, compliance and risk is the appropriate diagnostic because the observation—business case omits users reliant on assistive technologies—does not by itself establish every dependency, authorization rule or failure boundary. The findings explain the reported risk (project review scores optional chatbot convenience higher than accessibility and regulatory filing) before the design is approved.
Incorrect Answers
A:
The alternative “Prioritize accessible essential workflow before optional AI enhancements” is target architecture selection; it does not establish why the observed path failed. In the reported observation, business case omits users reliant on assistive technologies. The option “Translate user needs into measurable outcomes, compliance and risk” instead addresses the requested root-cause evidence.
C:
The alternative “Sequence the accessibility remediation with user testing and business ownership” is rollout work that may modify the wrong dependency before root-cause evidence is collected. In the reported observation, business case omits users reliant on assistive technologies. The option “Translate user needs into measurable outcomes, compliance and risk” instead addresses the requested root-cause evidence.
D:
Novelty does not meet the required filing accessibility outcome. For this question, it is not a defensible choice for root-cause evidence and it does not satisfy the requirement that benefit delivery must meet accessible filing requirements.
Question 867
The crew roster API environment at Regional airline faces this issue: a database primary restart causes regional crews to lose schedules for an entire shift. Supporting evidence shows that failure logs show read-only standby exists but application uses only one endpoint. The architect must preserve the requirement that dispatch must read the last approved crew plan during recovery. Select TWO complementary actions: one durable target architecture and one controlled way to introduce it while preserving service. Which TWO are best?
- Stage client reconnection and crew-cache refresh before database switchover
- Use high-availability database failover and a validated read-only contingency view
- Track replication lag, failover times and stale roster warnings
- Classify read/write operations, acceptable staleness and failover dependencies
- Point every client to an unreplicated local file as the only source
Explanation
Correct Answers
A:
Stage client reconnection and crew-cache refresh before database switchover is a controlled means to introduce the approved architecture while meeting the requirement that dispatch must read the last approved crew plan during recovery. Staged delivery, attributable change evidence and rollback guard against an otherwise avoidable service disruption. This supplies the controlled-introduction half of the two-action package.
B:
Use high-availability database failover and a validated read-only contingency view targets the underlying issue (a database primary restart causes regional crews to lose schedules for an entire shift) rather than merely reporting symptoms. It also preserves the required outcome (dispatch must read the last approved crew plan during recovery), which a target-state design must satisfy. This supplies the lasting-control half of the two-action package.
Incorrect Answers
C:
Monitoring crew roster API for future drift is useful after the rollout, but it does not specify the requested target control and safe implementation for this case: a database primary restart causes regional crews to lose schedules for an entire shift.
D:
This diagnoses the original issue (a database primary restart causes regional crews to lose schedules for an entire shift), but the TWO requested actions are the target architecture and its implementation; discovery was already completed.
E:
An uncontrolled copy can yield inconsistent safety-critical rosters. The shortcut does not supply the requested design and safe rollout for crew roster API, where failure logs show read-only standby exists but application uses only one endpoint.
Question 868
Global manufacturing supplier plans to update product catalog service. Investigators found that sample records have evolving nested attributes but no multirow transaction need, consistent with the reported problem that a document catalog changes shape frequently across suppliers and needs low-latency mobile reads. The response must satisfy the requirement that catalog availability matters more than cross-product SQL joins. The design review is over and production introduction is next. Which action is the best fit?
- Select Firestore document storage with explicit security rules and required indexes
- Make every supplier share an unrestricted project-wide data role
- Migrate one supplier feed and validate mobile access with rules emulator
- Track index growth, latency and document read costs
Explanation
Correct Answer
C:
Migrate one supplier feed and validate mobile access with rules emulator is a controlled means to introduce the approved architecture while meeting the requirement that catalog availability matters more than cross-product SQL joins. Staged delivery, attributable change evidence and rollback guard against an otherwise avoidable service disruption.
Incorrect Answers
A:
The alternative “Select Firestore document storage with explicit security rules and required indexes” describes the chosen design, but not how it is introduced safely. In the reported observation, sample records have evolving nested attributes but no multirow transaction need. The option “Migrate one supplier feed and validate mobile access with rules emulator” instead addresses the requested controlled implementation.
B:
Broad permissions negate document-level access rules. For this question, it is not a defensible choice for controlled implementation and it does not satisfy the requirement that catalog availability matters more than cross-product SQL joins.
D:
The alternative “Track index growth, latency and document read costs” addresses later monitoring rather than the actual controlled deployment. In the reported observation, sample records have evolving nested attributes but no multirow transaction need. The option “Migrate one supplier feed and validate mobile access with rules emulator” instead addresses the requested controlled implementation.
Question 869
Case study: Streaming fitness platform: event analytics ingestion
Streaming fitness platform relies on event analytics ingestion to deliver its operational commitments. The current architectural problem is that incoming device events arrive late and in duplicates and dashboards need windows. The team has recorded this evidence: sample payloads show device clocks skew ten minutes across regions. Any proposed response must preserve the requirement that analytics should not interfere with real-time device safety controls.
Streaming fitness platform plans to update event analytics ingestion. Investigators found that sample payloads show device clocks skew ten minutes across regions, consistent with the reported problem that incoming device events arrive late and in duplicates and dashboards need windows. The response must satisfy the requirement that analytics should not interfere with real-time device safety controls. The security and reliability teams dispute the root cause. Which evidence-gathering action is justified now?
- Use Pub/Sub with Dataflow event-time processing and BigQuery analytics sinks
- Roll out schema-versioned consumers with dead-letter handling
- Inspect event-time semantics, watermark lateness and replay handling
- Use request arrival timestamps as ground truth without clock handling
Explanation
Correct Answer
C:
Inspect event-time semantics, watermark lateness and replay handling is the appropriate diagnostic because the observation—sample payloads show device clocks skew ten minutes across regions—does not by itself establish every dependency, authorization rule or failure boundary. The findings explain the reported risk (incoming device events arrive late and in duplicates and dashboards need windows) before the design is approved.
Incorrect Answers
A:
The alternative “Use Pub/Sub with Dataflow event-time processing and BigQuery analytics sinks” is target architecture selection; it does not establish why the observed path failed. In the reported observation, sample payloads show device clocks skew ten minutes across regions. The option “Inspect event-time semantics, watermark lateness and replay handling” instead addresses the requested root-cause evidence.
B:
The alternative “Roll out schema-versioned consumers with dead-letter handling” is rollout work that may modify the wrong dependency before root-cause evidence is collected. In the reported observation, sample payloads show device clocks skew ten minutes across regions. The option “Inspect event-time semantics, watermark lateness and replay handling” instead addresses the requested root-cause evidence.
D:
Arrival time alone distorts late-event analytics. For this question, it is not a defensible choice for root-cause evidence and it does not satisfy the requirement that analytics should not interfere with real-time device safety controls.
Question 870
At International insurance portal, the automated rollback thresholds workload has a documented risk: new revision doubles checkout error rate but rollout continues because logs are not tied to release gates. The review found that service dashboard shows 6% failure during a 20% canary. The architecture must also ensure that customer billing changes must be reversible without inconsistent transactions. Select TWO complementary actions: one durable target architecture and one controlled way to introduce it while preserving service. Which TWO are best?
- Deploy gradual traffic with automated abort thresholds
- Treat release success as the number of pods that became Ready
- Track failed rollback events and release-specific SLO deltas
- Establish baseline error budgets, canary signals and rollback authority
- Define canary gates for error, latency and transaction integrity
Explanation
Correct Answers
A:
Deploy gradual traffic with automated abort thresholds is a controlled means to introduce the approved architecture while meeting the requirement that customer billing changes must be reversible without inconsistent transactions. Staged delivery, attributable change evidence and rollback guard against an otherwise avoidable service disruption. This supplies the controlled-introduction half of the two-action package.
E:
Define canary gates for error, latency and transaction integrity targets the underlying issue (new revision doubles checkout error rate but rollout continues because logs are not tied to release gates) rather than merely reporting symptoms. It also preserves the required outcome (customer billing changes must be reversible without inconsistent transactions), which a target-state design must satisfy. This supplies the lasting-control half of the two-action package.
Incorrect Answers
B:
Readiness alone does not demonstrate correct customer transactions. The shortcut does not supply the requested design and safe rollout for automated rollback thresholds, where service dashboard shows 6% failure during a 20% canary.
C:
Monitoring automated rollback thresholds for future drift is useful after the rollout, but it does not specify the requested target control and safe implementation for this case: new revision doubles checkout error rate but rollout continues because logs are not tied to release gates.
D:
This diagnoses the original issue (new revision doubles checkout error rate but rollout continues because logs are not tied to release gates), but the TWO requested actions are the target architecture and its implementation; discovery was already completed.