Google Professional Cloud Architect Practice Test 78
Question 770
Hospital payments administrator plans to update database point-in-time recovery. Investigators found that audit shows thousands of records changed before detection, consistent with the reported problem that an operator accidentally ran a broad UPDATE against the claims settlement database. The response must satisfy the requirement that finance requires restoration to immediately before the erroneous transaction. The team needs an operations signal rather than another design or one-time test. What should it track?
- Restore from a week-old cold snapshot and overwrite new valid payments blindly
- Set continuous backup/PITR where supported and segregated recovery permissions
- Restore to a pre-error timestamp and reconcile valid later transactions
- Track log-retention windows, backup gaps and restore test age
Explanation
Correct Answer
D:
Track log-retention windows, backup gaps and restore test age reveals recurrence, drift or changing dependencies after launch. It provides a repeatable signal for the risk (an operator accidentally ran a broad UPDATE against the claims settlement database) rather than treating the first successful test as permanent assurance.
Incorrect Answers
A:
An old snapshot without transaction reconciliation can discard legitimate payments. For this question, it is not a defensible choice for ongoing operational assurance and it does not satisfy the requirement that finance requires restoration to immediately before the erroneous transaction.
B:
The alternative “Set continuous backup/PITR where supported and segregated recovery permissions” selects a design, but does not provide the ongoing signal required to detect drift. In the reported observation, audit shows thousands of records changed before detection. The option “Track log-retention windows, backup gaps and restore test age” instead addresses the requested ongoing operational assurance.
C:
The alternative “Restore to a pre-error timestamp and reconcile valid later transactions” is a one-time acceptance result rather than ongoing operational assurance. In the reported observation, audit shows thousands of records changed before detection. The option “Track log-retention windows, backup gaps and restore test age” instead addresses the requested ongoing operational assurance.
Question 771
At National blood supply service, the database schema rollout workload has a documented risk: application instances migrate separately from the database schema and old pods fail on renamed columns. The review found that staging test shows older binary cannot read new tables. The architecture must also ensure that donation inventory cannot pause through rolling deployment. What follow-up metric or alert will identify emerging failures after the initial acceptance test?
- Drop the old column before the old application instances are drained
- Run old and new binaries against the transitional schema and reconcile inventory
- Monitor query failures, dead code paths and migration exceptions
- Define expand-migrate-contract schema evolution with compatible views
Explanation
Correct Answer
C:
Monitor query failures, dead code paths and migration exceptions reveals recurrence, drift or changing dependencies after launch. It provides a repeatable signal for the risk (application instances migrate separately from the database schema and old pods fail on renamed columns) rather than treating the first successful test as permanent assurance.
Incorrect Answers
A:
A destructive schema change strands older active binaries. For this question, it is not a defensible choice for ongoing operational assurance and it does not satisfy the requirement that donation inventory cannot pause through rolling deployment.
B:
The alternative “Run old and new binaries against the transitional schema and reconcile inventory” is a one-time acceptance result rather than ongoing operational assurance. In the reported observation, staging test shows older binary cannot read new tables. The option “Monitor query failures, dead code paths and migration exceptions” instead addresses the requested ongoing operational assurance.
D:
The alternative “Define expand-migrate-contract schema evolution with compatible views” selects a design, but does not provide the ongoing signal required to detect drift. In the reported observation, staging test shows older binary cannot read new tables. The option “Monitor query failures, dead code paths and migration exceptions” instead addresses the requested ongoing operational assurance.
Question 772
At Industrial robotics startup, the edge-device digital twin workload has a documented risk: architects assume GPU inference remains constant even though future robots gain local accelerators. The review found that hardware roadmap shows some plants will run most inference at the edge. The architecture must also ensure that plant safety systems must not depend on public network availability. The architecture review is paused until the team can prove the real failure path. Which investigation should happen first?
- Pilot one hardware generation with fallback and signed updates
- Forecast edge capability, model footprints and remote update channels
- Design split edge/cloud inference with versioned model contracts and offline safety path
- Force all safety-critical inference through a distant cloud endpoint
Explanation
Correct Answer
B:
Forecast edge capability, model footprints and remote update channels is the appropriate diagnostic because the observation—hardware roadmap shows some plants will run most inference at the edge—does not by itself establish every dependency, authorization rule or failure boundary. The findings explain the reported risk (architects assume GPU inference remains constant even though future robots gain local accelerators) before the design is approved.
Incorrect Answers
A:
The alternative “Pilot one hardware generation with fallback and signed updates” is rollout work that may modify the wrong dependency before root-cause evidence is collected. In the reported observation, hardware roadmap shows some plants will run most inference at the edge. The option “Forecast edge capability, model footprints and remote update channels” instead addresses the requested root-cause evidence.
C:
The alternative “Design split edge/cloud inference with versioned model contracts and offline safety path” is target architecture selection; it does not establish why the observed path failed. In the reported observation, hardware roadmap shows some plants will run most inference at the edge. The option “Forecast edge capability, model footprints and remote update channels” instead addresses the requested root-cause evidence.
D:
WAN outages would compromise real-time plant operation. For this question, it is not a defensible choice for root-cause evidence and it does not satisfy the requirement that plant safety systems must not depend on public network availability.
Question 773
The multi-project packet filtering environment at Education research foundation faces this issue: research projects inherit inconsistent firewall rules and centralized policy has no owner. Supporting evidence shows that network review finds SSH open to the internet from three experiment subnets. The architect must preserve the requirement that a limited set of approved remote researchers still needs audited access. What test must pass before the business accepts the remediated architecture?
- Inventory rule inheritance, exposure, identity and exception ownership
- Verify unapproved SSH is rejected while authenticated researcher routes remain
- Let every project administrator create a public any-any ingress rule
- Apply deny policies in monitored phases with approved access exceptions
Explanation
Correct Answer
B:
Verify unapproved SSH is rejected while authenticated researcher routes remain is a direct acceptance test of the path implicated by network review finds SSH open to the internet from three experiment subnets. It checks that the original risk (research projects inherit inconsistent firewall rules and centralized policy has no owner) has actually been controlled and tests that permitted operations still meet the requirement that a limited set of approved remote researchers still needs audited access.
Incorrect Answers
A:
The alternative “Inventory rule inheritance, exposure, identity and exception ownership” collects prechange evidence without confirming that the revised service meets its outcome. In the reported observation, network review finds SSH open to the internet from three experiment subnets. The option “Verify unapproved SSH is rejected while authenticated researcher routes remain” instead addresses the requested acceptance verification.
C:
Unrestricted local firewalls defeat centralized protection. For this question, it is not a defensible choice for acceptance verification and it does not satisfy the requirement that a limited set of approved remote researchers still needs audited access.
D:
The alternative “Apply deny policies in monitored phases with approved access exceptions” documents change execution but not its operational acceptance result. In the reported observation, network review finds SSH open to the internet from three experiment subnets. The option “Verify unapproved SSH is rejected while authenticated researcher routes remain” instead addresses the requested acceptance verification.
Question 774
Case study: National airport regulator: programmatic audit exports
National airport regulator relies on programmatic audit exports to deliver its operational commitments. The current architectural problem is that a Python client uses user ADC with project Owner and downloads more logs than needed. The team has recorded this evidence: code review finds broad filterless audit queries. Any proposed response must preserve the requirement that regulator requires a bounded report for one certified airport.
At National airport regulator, the Google Cloud design for programmatic audit exports is affected because a Python client uses user ADC with project Owner and downloads more logs than needed. A review found that code review finds broad filterless audit queries. The proposed response must account for the requirement that regulator requires a bounded report for one certified airport. Which analysis will allow the architect to distinguish the reported symptom from the actual failure path?
- Run unfiltered organization-wide log downloads with superadmin credentials
- Use scoped service identity and paginated filtered API requests
- Determine required log filters, export timestamps and reader identity
- Deploy export as an audited scheduled job with bounded destination
Explanation
Correct Answer
C:
Determine required log filters, export timestamps and reader identity is the appropriate diagnostic because the observation—code review finds broad filterless audit queries—does not by itself establish every dependency, authorization rule or failure boundary. The findings explain the reported risk (a Python client uses user ADC with project Owner and downloads more logs than needed) before the design is approved.
Incorrect Answers
A:
Overbroad access and exports exceed the audit's authorized scope. For this question, it is not a defensible choice for root-cause evidence and it does not satisfy the requirement that regulator requires a bounded report for one certified airport.
B:
The alternative “Use scoped service identity and paginated filtered API requests” is target architecture selection; it does not establish why the observed path failed. In the reported observation, code review finds broad filterless audit queries. The option “Determine required log filters, export timestamps and reader identity” instead addresses the requested root-cause evidence.
D:
The alternative “Deploy export as an audited scheduled job with bounded destination” is rollout work that may modify the wrong dependency before root-cause evidence is collected. In the reported observation, code review finds broad filterless audit queries. The option “Determine required log filters, export timestamps and reader identity” instead addresses the requested root-cause evidence.
Question 775
At E-commerce promotion service, the Google Cloud design for Cloud Run database connection storms is affected because a serverless revision scales quickly and exhausts a legacy SQL server's connection pool. A review found that database logs show connection refusals at promotion launch. The proposed response must account for the requirement that backend must remain within database concurrency limits. The solution is operating. Which evidence should be watched for longer-term reliability and compliance?
- Allow unlimited cold-start instances to open independent large pools
- Set Cloud Run concurrency and instance limits with appropriate pooling
- Track rejected requests, instance count and DB pool utilization
- Generate realistic parallel traffic and verify connections stay within ceiling
Explanation
Correct Answer
C:
Track rejected requests, instance count and DB pool utilization reveals recurrence, drift or changing dependencies after launch. It provides a repeatable signal for the risk (a serverless revision scales quickly and exhausts a legacy SQL server's connection pool) rather than treating the first successful test as permanent assurance.
Incorrect Answers
A:
Unbounded scaling overwhelms a fixed database dependency. For this question, it is not a defensible choice for ongoing operational assurance and it does not satisfy the requirement that backend must remain within database concurrency limits.
B:
The alternative “Set Cloud Run concurrency and instance limits with appropriate pooling” selects a design, but does not provide the ongoing signal required to detect drift. In the reported observation, database logs show connection refusals at promotion launch. The option “Track rejected requests, instance count and DB pool utilization” instead addresses the requested ongoing operational assurance.
D:
The alternative “Generate realistic parallel traffic and verify connections stay within ceiling” is a one-time acceptance result rather than ongoing operational assurance. In the reported observation, database logs show connection refusals at promotion launch. The option “Track rejected requests, instance count and DB pool utilization” instead addresses the requested ongoing operational assurance.
Question 776
Retail healthcare supplier plans to update Cloud Run canary release. Investigators found that release journal shows no canary or rollback criterion, consistent with the reported problem that a monolithic new API revision receives 100% of traffic and breaks patient orders. The response must satisfy the requirement that approved orders must continue during application upgrades. The security and reliability teams dispute the root cause. Which evidence-gathering action is justified now?
- Replace the production revision all at once and delete the prior image
- Analyze failure budget, revision behavior and traffic splitting needs
- Shift a small traffic percentage and widen only when health gates pass
- Design progressive Cloud Run revisions with measurable abort conditions
Explanation
Correct Answer
B:
Analyze failure budget, revision behavior and traffic splitting needs is the appropriate diagnostic because the observation—release journal shows no canary or rollback criterion—does not by itself establish every dependency, authorization rule or failure boundary. The findings explain the reported risk (a monolithic new API revision receives 100% of traffic and breaks patient orders) before the design is approved.
Incorrect Answers
A:
An irreversible full cutover magnifies deployment failure. For this question, it is not a defensible choice for root-cause evidence and it does not satisfy the requirement that approved orders must continue during application upgrades.
C:
The alternative “Shift a small traffic percentage and widen only when health gates pass” is rollout work that may modify the wrong dependency before root-cause evidence is collected. In the reported observation, release journal shows no canary or rollback criterion. The option “Analyze failure budget, revision behavior and traffic splitting needs” instead addresses the requested root-cause evidence.
D:
The alternative “Design progressive Cloud Run revisions with measurable abort conditions” is target architecture selection; it does not establish why the observed path failed. In the reported observation, release journal shows no canary or rollback criterion. The option “Analyze failure budget, revision behavior and traffic splitting needs” instead addresses the requested root-cause evidence.
Question 777
The Google Cloud SDK workload identity environment at Telecom customer portal faces this issue: a deployment script has a static JSON key checked into a Git repository. Supporting evidence shows that commit scan reveals key reuse across three build agents. The architect must preserve the requirement that deployments need constrained short-lived credentials. Before approving a solution, which analysis most directly clarifies the underlying dependency or control gap?
- Inventory CI identity provider claims and project permissions
- Use Workload Identity Federation with constrained provider attributes
- Copy the static key into the CI variable store without revoking it
- Transition build jobs to federation and revoke leaked keys
Explanation
Correct Answer
A:
Inventory CI identity provider claims and project permissions is the appropriate diagnostic because the observation—commit scan reveals key reuse across three build agents—does not by itself establish every dependency, authorization rule or failure boundary. The findings explain the reported risk (a deployment script has a static JSON key checked into a Git repository) before the design is approved.
Incorrect Answers
B:
The alternative “Use Workload Identity Federation with constrained provider attributes” is target architecture selection; it does not establish why the observed path failed. In the reported observation, commit scan reveals key reuse across three build agents. The option “Inventory CI identity provider claims and project permissions” instead addresses the requested root-cause evidence.
C:
Relocating the leaked credential does not eliminate its misuse risk. For this question, it is not a defensible choice for root-cause evidence and it does not satisfy the requirement that deployments need constrained short-lived credentials.
D:
The alternative “Transition build jobs to federation and revoke leaked keys” is rollout work that may modify the wrong dependency before root-cause evidence is collected. In the reported observation, commit scan reveals key reuse across three build agents. The option “Inventory CI identity provider claims and project permissions” instead addresses the requested root-cause evidence.
Question 778
Case study: International bank holding company: regulated data-center extension
International bank holding company relies on regulated data-center extension to deliver its operational commitments. The current architectural problem is that the proposed Dedicated Interconnect lacks a backup location and depends on a single metro facility. The team has recorded this evidence: site survey lists one colocation provider and a common power feed for both links. Any proposed response must preserve the requirement that settlement services require redundant connectivity with predictable private latency.
International bank holding company plans to update regulated data-center extension. Investigators found that site survey lists one colocation provider and a common power feed for both links, consistent with the reported problem that the proposed Dedicated Interconnect lacks a backup location and depends on a single metro facility. The response must satisfy the requirement that settlement services require redundant connectivity with predictable private latency. The security and reliability teams dispute the root cause. Which evidence-gathering action is justified now?
- Design geographically separated Interconnect circuits with independent BGP sessions and failover
- Install two ports in the same rack and call them regionally redundant
- Coordinate circuit turn-up and route preference in staged change windows
- Analyze metro failure domains, routing owners and throughput commitments
Explanation
Correct Answer
D:
Analyze metro failure domains, routing owners and throughput commitments is the appropriate diagnostic because the observation—site survey lists one colocation provider and a common power feed for both links—does not by itself establish every dependency, authorization rule or failure boundary. The findings explain the reported risk (the proposed Dedicated Interconnect lacks a backup location and depends on a single metro facility) before the design is approved.
Incorrect Answers
A:
The alternative “Design geographically separated Interconnect circuits with independent BGP sessions and failover” is target architecture selection; it does not establish why the observed path failed. In the reported observation, site survey lists one colocation provider and a common power feed for both links. The option “Analyze metro failure domains, routing owners and throughput commitments” instead addresses the requested root-cause evidence.
B:
Common power and facility dependencies defeat geographic resilience. For this question, it is not a defensible choice for root-cause evidence and it does not satisfy the requirement that settlement services require redundant connectivity with predictable private latency.
C:
The alternative “Coordinate circuit turn-up and route preference in staged change windows” is rollout work that may modify the wrong dependency before root-cause evidence is collected. In the reported observation, site survey lists one colocation provider and a common power feed for both links. The option “Analyze metro failure domains, routing owners and throughput commitments” instead addresses the requested root-cause evidence.
Question 779
Case study: Healthcare billing analytics: log-based incident metrics
Healthcare billing analytics relies on log-based incident metrics to deliver its operational commitments. The current architectural problem is that an API error log is present but there is no alert on rising failed invoice reconciliation. The team has recorded this evidence: log query shows 15% payment finalization errors during partner outage. Any proposed response must preserve the requirement that financial integrity requires detection before end-of-day settlement.
At Healthcare billing analytics, the log-based incident metrics workload has a documented risk: an API error log is present but there is no alert on rising failed invoice reconciliation. The review found that log query shows 15% payment finalization errors during partner outage. The architecture must also ensure that financial integrity requires detection before end-of-day settlement. The architecture review is paused until the team can prove the real failure path. Which investigation should happen first?
- Identify a reliable error signature, baseline and affected invoice identifiers
- Only send a weekly email digest for critical reconciliation failures
- Define a log-based metric and threshold tied to reconciliation impact
- Introduce the metric and alert route with synthetic nonfinancial test events
Explanation
Correct Answer
A:
Identify a reliable error signature, baseline and affected invoice identifiers is the appropriate diagnostic because the observation—log query shows 15% payment finalization errors during partner outage—does not by itself establish every dependency, authorization rule or failure boundary. The findings explain the reported risk (an API error log is present but there is no alert on rising failed invoice reconciliation) before the design is approved.
Incorrect Answers
B:
A weekly summary misses urgent financial incidents. For this question, it is not a defensible choice for root-cause evidence and it does not satisfy the requirement that financial integrity requires detection before end-of-day settlement.
C:
The alternative “Define a log-based metric and threshold tied to reconciliation impact” is target architecture selection; it does not establish why the observed path failed. In the reported observation, log query shows 15% payment finalization errors during partner outage. The option “Identify a reliable error signature, baseline and affected invoice identifiers” instead addresses the requested root-cause evidence.
D:
The alternative “Introduce the metric and alert route with synthetic nonfinancial test events” is rollout work that may modify the wrong dependency before root-cause evidence is collected. In the reported observation, log query shows 15% payment finalization errors during partner outage. The option “Identify a reliable error signature, baseline and affected invoice identifiers” instead addresses the requested root-cause evidence.
Question 780
The short-lived GPU inference environment at AI satellite imagery group faces this issue: the product team leaves GPU instances running overnight for occasional image classification. Supporting evidence shows that billing shows high GPU idle cost with short bursts of queued requests. The architect must preserve the requirement that inference can tolerate several seconds of cold-start delay. What test must pass before the business accepts the remediated architecture?
- Canary burst routing with model artifact caching and capped quotas
- Verify output equivalence through GPU scale-down and restart events
- Measure model footprint, arrival patterns and accelerator warm-up cost
- Reserve the largest GPU fleet permanently for infrequent requests
Explanation
Correct Answer
B:
Verify output equivalence through GPU scale-down and restart events is a direct acceptance test of the path implicated by billing shows high GPU idle cost with short bursts of queued requests. It checks that the original risk (the product team leaves GPU instances running overnight for occasional image classification) has actually been controlled and tests that permitted operations still meet the requirement that inference can tolerate several seconds of cold-start delay.
Incorrect Answers
A:
The alternative “Canary burst routing with model artifact caching and capped quotas” documents change execution but not its operational acceptance result. In the reported observation, billing shows high GPU idle cost with short bursts of queued requests. The option “Verify output equivalence through GPU scale-down and restart events” instead addresses the requested acceptance verification.
C:
The alternative “Measure model footprint, arrival patterns and accelerator warm-up cost” collects prechange evidence without confirming that the revised service meets its outcome. In the reported observation, billing shows high GPU idle cost with short bursts of queued requests. The option “Verify output equivalence through GPU scale-down and restart events” instead addresses the requested acceptance verification.
D:
Always-on capacity wastes cost when demand is highly intermittent. For this question, it is not a defensible choice for acceptance verification and it does not satisfy the requirement that inference can tolerate several seconds of cold-start delay.