Courseiva

CCNA Scaling With Google Cloud Operations Questions

8 of 83 questions · Page 2/2 · Scaling With Google Cloud Operations topic · Answers revealed

76
MCQhard

A cloud operations engineer notices that the managed instance group 'my-mig' has been scaling up frequently, but the application performance is still degraded. The CPU utilization metric shows high values. What is most likely the issue?

A.The target size is set to 10, which is lower than the current needed capacity.
B.The instance group is using preemptible VMs which are being reclaimed frequently.
C.The autoscaler is using a cooldown period that is too long, preventing it from scaling down.
D.The scaling metric is not appropriate; consider using a custom metric that better reflects application load.
AnswerD

The autoscaler is scaling based on a metric that does not accurately represent the application's real load, which is why performance remains degraded even as instances increase. CPU utilization is often a poor proxy for managed instance groups because workloads can be I/O-bound, memory-bound, or dependent on external queues and services. A custom metric such as request latency, queue depth, or concurrent requests directly measures the workload bottleneck and is recommended for autoscaling policies. Switching to a metric that aligns with the application's actual performance signals would enable the autoscaler to make precise scaling decisions and maintain target service levels.

Why this answer

The autoscaler is using CPU utilization as the scaling metric, but high CPU does not necessarily correlate with application performance degradation. If the application is bottlenecked on memory, I/O, or request queuing, CPU may remain high while throughput suffers. A custom metric (e.g., requests per second, latency, or queue depth) would better reflect actual application load and enable more accurate scaling decisions.

Exam trap

The trap here is that candidates assume high CPU utilization always means the application needs more compute capacity, but the question tests the understanding that the scaling metric must be aligned with the actual performance bottleneck, not just a generic system metric.

How to eliminate wrong answers

Option A is wrong because the target size being lower than needed capacity would prevent scaling up sufficiently, but the question states the instance group is scaling up frequently, so the autoscaler is actively adding instances; the issue is that scaling up is not fixing the performance problem. Option B is wrong because preemptible VMs being reclaimed would cause instance churn and potential performance degradation, but the question does not mention preemptible VMs, and the symptom of frequent scaling up with high CPU is not directly caused by preemption. Option C is wrong because a cooldown period that is too long would delay scaling down, not prevent scaling up; the issue here is that scaling up is happening but not resolving the degradation, so the cooldown period is not the root cause.

77
MCQmedium

Refer to the exhibit. A DevOps engineer wants to create a chart showing the rate of items sold per second over time. What is a limitation of this metric for that purpose?

A.The metric kind is GAUGE, so it cannot be used to calculate rate
B.The interval should include a startTime
C.The metric has no labels to filter
D.The value should be DOUBLE instead of INT64
AnswerA

GAUGE metrics capture a value at a single point in time, like a snapshot, so there is no temporal difference to divide by an elapsed interval. Cloud Monitoring's rate aligner (ALIGN_RATE) or the 'rate' transform requires DELTA or CUMULATIVE metrics to compute per-second change. Because the metric is GAUGE, it is structurally impossible to calculate a rate, regardless of the interval or value type.

Why this answer

A GAUGE metric type represents a point-in-time value (e.g., current number of items), not a cumulative counter. To calculate a rate (items per second), you need a CUMULATIVE counter metric that monotonically increases, allowing Cloud Monitoring to compute the derivative over time. GAUGE metrics lack the necessary monotonicity and cumulative semantics, so they cannot be used to derive a meaningful rate of change.

Exam trap

Google Cloud often tests the misconception that any numeric metric can be used to compute a rate, when in fact only CUMULATIVE counters support rate-of-change calculations in Cloud Monitoring.

How to eliminate wrong answers

Option B is wrong because including a startTime in the interval is not a limitation of the metric itself; it is a standard parameter for time-series queries and does not prevent rate calculation. Option C is wrong because the absence of labels does not prevent rate calculation; labels are for filtering and aggregation, not for the fundamental ability to compute a rate. Option D is wrong because the data type (INT64 vs DOUBLE) does not affect the ability to calculate a rate; Cloud Monitoring can compute rates on integer values, and the limitation is the metric kind (GAUGE vs CUMULATIVE), not the value type.

78
MCQmedium

A company's cloud operations team is implementing a tagging strategy for cost allocation. They want to ensure that the 'cost-center' label is present on every Compute Engine VM and Cloud Storage bucket created in their Google Cloud organization. Currently, some resources are created without this label. Which combination of controls best enforces and remediates this requirement?

A.Organization Policy custom constraint to prevent creation of resources without the 'cost-center' label (preventive), plus Cloud Asset Inventory to identify existing unlabeled resources for remediation (detective)
B.Only organization policy — once new resources are blocked, existing unlabeled resources don't matter
C.Only Cloud Asset Inventory monitoring — alerting on unlabeled resources is sufficient without preventing their creation
D.Grant all engineers the 'Labels Admin' role to encourage them to add labels voluntarily
AnswerA

This is the complete two-layer approach: prevention (org policy blocks future non-compliant resources at creation time) and detection/remediation (Cloud Asset Inventory finds existing unlabeled resources so they can be labeled retroactively). Together they address both the future and existing state.

Why this answer

A preventive control (org policy custom constraint requiring the label) stops future non-compliant resources. A detective/corrective control (Cloud Asset Inventory + Cloud Functions or Security Command Center) finds and remediates existing unlabeled resources. Both are needed for comprehensive enforcement.

79
MCQmedium

After a major production outage, the engineering team conducts a review of what happened, why it happened, and how to prevent it in the future. This document is shared with all engineering teams. What is this practice called, and why does Google's SRE culture emphasize it?

A.Performance review — identifying which engineers caused the outage for disciplinary action.
B.Blameless postmortem — documenting the incident, root causes, and preventive actions to drive systemic learning without individual blame.
C.Capacity planning review — ensuring enough servers are provisioned to prevent future outages.
D.Change advisory board (CAB) review — approving that the outage fix is safe to deploy.
AnswerB

A blameless postmortem is a retrospective, collaborative analysis of an incident that produces a written record containing a detailed timeline, impact assessment, root cause analysis (often using techniques such as '5 Whys' or cause-effect diagrams), and concrete preventive actions. The 'blameless' attribute ensures the investigation concentrates on systemic contributions — like incomplete monitoring, ambiguous runbooks, or brittle configurations — rather than punishing the operator who triggered the failure. This approach builds psychological safety, enabling honest reporting of contributing factors, and turns failures into opportunities for continuous improvement, which is a fundamental pillar of reliability engineering.

Why this answer

A blameless postmortem is a core SRE practice that focuses on documenting incidents, root causes, and preventive actions without assigning individual blame. Google's SRE culture emphasizes this to foster psychological safety, enabling teams to openly share failures and drive systemic improvements, which is essential for maintaining high reliability in large-scale distributed systems.

Exam trap

The trap here is that candidates may confuse a blameless postmortem with a performance review or a change management process, failing to recognize that the key differentiator is the absence of blame and the focus on systemic learning rather than individual accountability.

How to eliminate wrong answers

Option A is wrong because a performance review is an HR process for evaluating employee contributions, not a post-incident analysis; blaming individuals contradicts the blameless culture that encourages honest incident reporting. Option C is wrong because capacity planning review is a proactive process to ensure sufficient resources (e.g., servers, network bandwidth) are provisioned to meet demand, not a reactive review of a specific outage's causes and fixes. Option D is wrong because a change advisory board (CAB) review is an ITIL process for approving changes before deployment, not a retrospective analysis of an incident that has already occurred.

80
MCQhard

A company wants to implement SLOs for their API service. They need to measure the proportion of successful requests over a 30-day window. Which metric should they use?

A.availability (uptime)
B.latency at 99th percentile
C.requests/success
D.SLI = good events / total events
AnswerD

An SLI defined as good events divided by total events is the canonical, user-centric approach for an API service, as it directly measures the proportion of requests that meet predefined success criteria. For example, a request may be considered 'good' if it returns an HTTP 2xx status and has a latency below a threshold, while 'total events' counts every request received. This ratio yields a percentage that aligns with the actual user experience and forms the basis for setting an SLO and calculating an error budget, following standard SRE practices.

Why this answer

An SLI (Service Level Indicator) is defined as the ratio of good events to total events, which directly measures the proportion of successful requests over a 30-day window. This aligns with the requirement to track request success rate, not just system uptime. In Google Cloud operations, SLOs are built on SLIs that count discrete events like HTTP 200 responses versus all requests.

Exam trap

The trap here is that candidates confuse availability (uptime) with request success rate, not realizing that a service can be 'up' 100% of the time yet fail a large proportion of requests due to application errors.

How to eliminate wrong answers

Option A is wrong because availability (uptime) measures the percentage of time the service is reachable, not the proportion of individual request successes; a service can be up but still return errors for many requests. Option B is wrong because latency at the 99th percentile measures response time distribution, not success rate; it addresses performance, not correctness or error rate. Option C is wrong because requests/success is an inverted ratio that would decrease as success increases, and it is not a standard SLI formula; the correct SLI is good events divided by total events.

81
MCQhard

An operations team tracks the following metrics for their customer portal: request latency p99, error rate, and requests per second. In Site Reliability Engineering terminology, what are these metrics called, and what do they collectively define?

A.Key Performance Indicators (KPIs) that define the overall health of the business
B.Service Level Agreements (SLAs), defining the contractual commitments made to customers
C.Service Level Indicators (SLIs), which measure specific dimensions of service behavior from the user's perspective and collectively define how reliability is quantified
D.Operational metrics that are only relevant to the infrastructure team and not to business stakeholders
AnswerC

SLIs are the specific measurable quantities that capture how users experience the service. Latency (is it fast enough?), error rate (is it working?), and throughput (is it keeping up?) are the canonical SLI types. Together they provide a quantitative picture of reliability that can be used to set SLO targets.

Why this answer

In Site Reliability Engineering (SRE), the metrics p99 latency, error rate, and requests per second are classified as Service Level Indicators (SLIs). SLIs are carefully chosen quantitative measures of specific aspects of the service's behavior, such as availability, latency, or throughput, as experienced by the end user. Collectively, these SLIs define how reliability is quantified and are used to set and monitor Service Level Objectives (SLOs).

Exam trap

The trap here is that candidates confuse SLIs with SLAs or KPIs, not realizing that SLIs are the raw measurements that feed into SLOs, which then underpin SLAs, and that they are specifically defined from the user's perspective to quantify reliability.

How to eliminate wrong answers

Option A is wrong because while these metrics can be part of business KPIs, the SRE terminology specifically calls them Service Level Indicators (SLIs), not generic KPIs, and they define reliability quantification, not overall business health. Option B is wrong because SLAs are contractual commitments based on SLOs, which are in turn derived from SLIs; the metrics themselves are not the agreements. Option D is wrong because SLIs are explicitly defined from the user's perspective and are critical for business stakeholders to understand service reliability, not just for the infrastructure team.

82
MCQeasy

A company's production database is running on a Compute Engine VM with a 500 GB Persistent Disk. The operations team wants to create a backup they can restore from in case of data corruption or accidental deletion. Which Google Cloud capability provides point-in-time backup for Persistent Disks?

A.Cloud Storage bucket replication, by continuously copying the database files to a storage bucket
B.Persistent Disk Snapshots, which capture the disk state at a point in time and enable restoration or creation of new disks from that snapshot
C.Cloud SQL automated backups, which protect databases running on Compute Engine VMs
D.VM live migration, which moves the running VM between physical hosts, automatically creating a backup in the process
AnswerB

Persistent Disk Snapshots are the correct mechanism. They capture a consistent point-in-time image of the disk (application-consistent when used with snapshot agent or after flushing I/O). Snapshots are stored in Cloud Storage, incremental after the first snapshot, and can be used to create a new disk or restore data.

Why this answer

Persistent Disk Snapshots are the correct Google Cloud feature for creating point-in-time backups of Persistent Disks. They capture the disk's data and configuration at a specific moment, allowing you to restore the disk or create new disks from that snapshot. This is the native, recommended method for backup and disaster recovery of Compute Engine VM disks.

Exam trap

The trap here is that candidates confuse Cloud SQL backups (which are for managed databases) with the need to back up a database running on a Compute Engine VM, leading them to select option C instead of the correct Persistent Disk Snapshots.

How to eliminate wrong answers

Option A is wrong because Cloud Storage bucket replication is a feature for objects in buckets, not for Persistent Disks; continuously copying database files to a bucket would require custom scripting and does not provide crash-consistent point-in-time backups of the entire disk. Option C is wrong because Cloud SQL automated backups protect Cloud SQL managed databases, not databases running on Compute Engine VMs; Cloud SQL is a separate managed service, not a feature for Compute Engine disks. Option D is wrong because VM live migration moves a running VM between physical hosts for maintenance without downtime, but it does not create a backup or capture a point-in-time state of the disk.

83
MCQmedium

A company wants to set up automated checks that continuously verify their website's homepage, login page, and API endpoints are accessible from multiple global locations. If any endpoint becomes unreachable for more than 2 minutes, the on-call engineer should be alerted. Which Cloud Monitoring feature provides this?

A.Cloud Logging log-based alerts that detect 5xx errors in application logs.
B.Cloud Monitoring uptime checks that probe endpoints from global locations with alerting on failure.
C.Cloud Trace that records response times for each user request.
D.Custom scripts on Compute Engine VMs that ping endpoints every minute.
AnswerB

Cloud Monitoring uptime checks are a managed, synthetic monitoring solution that sends HTTP, HTTPS, or TCP probe requests to your endpoints from multiple global points of presence (PoPs) at configurable intervals. Alerting policies can be triggered when a configurable number of locations report failure, which verifies external availability from the user's perspective, not just internal health. This is the correct choice because it is a deliberately designed, low-maintenance service for exactly this use case, requiring no custom infrastructure and providing global coverage by default.

Why this answer

Cloud Monitoring uptime checks are specifically designed to probe HTTP, HTTPS, or TCP endpoints from multiple global locations at configurable intervals (e.g., every 1 minute). They can trigger alerting policies when a check fails for a specified duration (e.g., 2 minutes), directly matching the requirement for continuous, multi-location endpoint accessibility verification with alerting on sustained failure.

Exam trap

The trap here is that candidates confuse log-based alerts (which detect errors in logs) with proactive uptime checks (which test connectivity), leading them to choose Option A because they think 5xx errors are the only way to detect unreachability, ignoring that a completely down endpoint may not generate logs at all.

How to eliminate wrong answers

Option A is wrong because Cloud Logging log-based alerts analyze log entries (e.g., 5xx errors) but do not actively probe endpoints from global locations; they react to logs already generated, not to connectivity failures that may not produce logs. Option C is wrong because Cloud Trace is a distributed tracing tool that captures latency and request paths for individual user requests, not a monitoring feature for endpoint availability from multiple locations. Option D is wrong because custom scripts on Compute Engine VMs would require manual setup, lack native multi-location probing, and do not integrate with Cloud Monitoring's alerting policies; they are an ad-hoc solution, not a managed service.

← PreviousPage 2 of 2 · 83 questions total

Ready to test yourself?

Try a timed practice session using only Scaling With Google Cloud Operations questions.