SaveMyCert
Log in
5 of 5 free questions left today·for 10 a day
SOA-C03 · Domain 1

Monitoring, Logging, Analysis, Remediation, and Performance Optimization practice questions

Monitoring, Logging, Analysis, Remediation, and Performance Optimization is worth 22% of the SOA-C03 exam — the heaviest of the 5 domains. CloudWatch-centered observability — metrics, alarms, and logs — plus automated remediation and performance tuning for compute, storage, and databases. 6 fully worked examples are further down this page, answers included.

Exam weight
22%
the heaviest of the 5 domains
Questions
114
across 3 topics
Free, no account
5/day
a free account allows 10 a day
Explanations
Every option
right and wrong

Build a practice session

5 free questions left today.

Domains

How many?

5 free questions left today.

Mode

Ready when you are

10 fresh questions drawn across 1 of 5 domains, in Learn mode.

Review what you missed

Every question you answer incorrectly, and every question you flag while practising, is saved here automatically. Finish a session and you can come back to re-drill just those.

6 sample Monitoring, Logging, Analysis, Remediation, and Performance Optimization questions, fully explained

Questions from the SOA-C03 bank mapped to domain 1, with the answer key and the reasoning behind every option. None of them repeat the examples on the main SOA-C03 practice page.

Question 1Monitoring, Logging, Analysis, Remediation, and Performance Optimization

A CloudOps team manages 300 running EC2 instances that are already Systems Manager managed nodes. The team must start collecting memory and disk-space metrics. Any later change to the collected metrics must be made in one place and reach every running instance without replacing or rebooting it. Which approach meets these requirements with the LEAST operational overhead?

Choose one.

  • a
    Store the agent configuration in Parameter Store and apply it with the AmazonCloudWatch-ManageAgent document. Correct

    A single configuration stored in Parameter Store is the one source of truth, and Run Command with AmazonCloudWatch-ManageAgent (after AWS-ConfigureAWSPackage installs the agent) pushes it to every running managed node; a later change is a parameter update plus one more command run.

  • b
    Run AWS-ConfigureAWSPackage through a State Manager association that targets the fleet's managed nodes.

    This installs and keeps the CloudWatch agent package current on every node, but the document applies no agent configuration, so nothing tells the agent to collect memory or disk metrics.

  • c
    Keep one agent configuration file in an S3 bucket and have each instance's boot script download it.

    There is one copy of the configuration, but instances read it only when they boot, so the running fleet never picks it up and every later change needs a reboot of all 300 instances.

  • d
    Put the agent install and configuration file in the user data of the launch templates the fleet uses.

    User data runs at first boot of new instances only, so the 300 running instances are not configured and later changes again require replacing instances.

The concept

The CloudWatch agent reads a JSON configuration. Storing that configuration in Systems Manager Parameter Store and applying it with Systems Manager (AWS-ConfigureAWSPackage to install, AmazonCloudWatch-ManageAgent to configure) gives a whole fleet one consistent, centrally changed configuration.

Why that’s the answer

Three constraints decide it: the instances are already running, they must not be replaced or rebooted, and later changes must be made in one place. Parameter Store plus the AmazonCloudWatch-ManageAgent document meets all three; re-running the command after editing the parameter updates every node in place. AWS-ConfigureAWSPackage is the near-twin: it installs the agent but does not configure it. A configuration file read at boot and launch-template user data only take effect when an instance starts.

How to reason it out
  1. Note that in-guest metrics (memory, disk space) need the CloudWatch agent with a configuration that lists them.
  2. Note the constraints: running instances, no replacement or reboot, one place for later changes.
  3. Eliminate boot-time mechanisms (user data, a file downloaded at startup) because running instances never read them.
  4. Eliminate AWS-ConfigureAWSPackage on its own because it installs the agent but applies no configuration.
  5. Choose Parameter Store for the configuration and AmazonCloudWatch-ManageAgent to apply it.

Exam tip: Fleet-wide CloudWatch agent: configuration in Parameter Store, installed and applied with Systems Manager.

CloudWatch Metrics, Alarms, and Log Filters for SOA-C03 — the lesson that teaches this.

Question 2Monitoring, Logging, Analysis, Remediation, and Performance Optimization

An Auto Scaling group scales on average CPUUtilization, but scaling reacts slowly because the EC2 instance metrics arrive only every 5 minutes. What is the MOST direct way to make the instance metrics available at 1-minute frequency?

Choose one.

  • a
    Install the CloudWatch agent on the instances.

    The agent's job is collecting in-guest metrics such as memory and log files. The frequency of the standard CPUUtilization metric is controlled by detailed monitoring, not the agent.

  • b
    Publish CPU usage as a high-resolution custom metric with PutMetricData.

    This would require writing and maintaining custom collection code when a built-in setting already delivers 1-minute CPUUtilization.

  • c
    Enable detailed monitoring on the instances. Correct

    Detailed monitoring raises EC2 instance metrics from the free 5-minute standard frequency to 1-minute frequency for an additional cost — exactly what faster scaling reactions need.

  • d
    Shorten the alarm period on the scaling policy to 1 minute.

    An alarm cannot evaluate faster than datapoints arrive. With 5-minute data, 1-minute periods just produce missing datapoints.

The concept

EC2 standard monitoring publishes instance metrics every 5 minutes at no charge. Detailed monitoring publishes the same metrics every 1 minute for an additional cost — it changes frequency only, adding no new metrics — and is enabled per instance or in a launch template.

Why that’s the answer

Enabling detailed monitoring is the built-in, direct way to get 1-minute EC2 metrics so alarms and Auto Scaling react faster. The CloudWatch agent solves a different problem (in-guest metrics and logs); a high-resolution custom metric means writing collection code for data AWS already provides; and shrinking the alarm period cannot conjure datapoints that are only published every 5 minutes.

How to reason it out
  1. Identify that the bottleneck is metric frequency, not metric availability.
  2. Enable detailed monitoring on the instances or in their launch template.
  3. Confirm CPUUtilization now arrives at 1-minute granularity.
  4. Adjust the scaling policy's alarm period to evaluate 1-minute datapoints.

Exam tip: Standard monitoring is 5-minute and free; detailed monitoring is 1-minute and paid — and it changes frequency, never the metric set.

CloudWatch Metrics, Alarms, and Log Filters for SOA-C03 — the lesson that teaches this.

Question 3Monitoring, Logging, Analysis, Remediation, and Performance Optimization

An application's internal queue depth can spike and cause failures within about 30 seconds. The team publishes the queue depth with PutMetricData and needs an alarm that can detect a spike and notify the on-call well under a minute after it starts. Which combination of publishing and alarm settings meets the requirement?

Choose one.

  • a
    Publish with StorageResolution set to 60 and create an alarm with a 10-second period.

    A StorageResolution of 60 stores the metric at 1-minute granularity, so a 10-second alarm period has no sub-minute datapoints to evaluate; high-resolution periods need a high-resolution metric.

  • b
    Publish with StorageResolution set to 1 and create an alarm with a 60-second period.

    The metric is stored at 1-second granularity, but a 60-second alarm period aggregates it into one datapoint a minute, so detection can take a minute or more.

  • c
    Publish with StorageResolution set to 60 and create an alarm with a 60-second period.

    This is the standard-resolution default: 1-minute data evaluated once a minute, which cannot react well under a minute.

  • d
    Publish with StorageResolution set to 1 and create an alarm with a 10-second period. Correct

    A StorageResolution of 1 makes the custom metric high resolution, and only a high-resolution metric supports a high-resolution alarm with a 10-second (or 30-second) period, giving sub-minute detection.

The concept

Custom metrics are standard resolution (StorageResolution 60, 1-minute granularity) by default or high resolution (StorageResolution 1, 1-second granularity). Alarms on high-resolution metrics can use 10-second or 30-second periods; any other alarm evaluates in multiples of 60 seconds.

Why that’s the answer

Sub-minute detection needs both halves: a high-resolution metric and a high-resolution alarm period. Publishing at 1-second resolution but alarming every 60 seconds wastes the resolution, and a 10-second alarm on 1-minute data has nothing to evaluate between minutes. High-resolution alarms cost more than standard ones, so reserve them for metrics that gate paging or scaling.

How to reason it out
  1. Translate the requirement: detection well under 60 seconds.
  2. Check the publishing side: only StorageResolution 1 stores sub-minute datapoints.
  3. Check the alarm side: only a 10- or 30-second period evaluates sub-minute.
  4. Pick the option where both settings are high resolution.

Exam tip: Sub-minute alarms need a high-resolution metric (StorageResolution 1) and a high-resolution alarm period (10 or 30 seconds).

CloudWatch Metrics, Alarms, and Log Filters for SOA-C03 — the lesson that teaches this.

Question 4Monitoring, Logging, Analysis, Remediation, and Performance Optimization

A team runs microservices on an Amazon ECS cluster whose tasks all use AWS Fargate. The team wants CPU and memory metrics at cluster, service and task level in CloudWatch, plus the performance data behind them, without writing collection code. Which option meets the requirement with the LEAST operational effort?

Choose one.

  • a
    Add a CloudWatch agent sidecar container to the task definitions of the cluster's services.

    A sidecar agent does run on Fargate, but every task definition must be edited and every task carries an extra container to configure and maintain, which is more effort than a cluster setting.

  • b
    Scrape the tasks with a managed collector and store the metrics in Managed Service for Prometheus.

    This stores Prometheus-format metrics outside CloudWatch and needs exporters in the services, so it does not deliver ready-made ECS metrics and logs in CloudWatch.

  • c
    Send task logs with the awslogs driver and build metric filters for CPU and memory usage.

    The awslogs driver does ship container logs, but application logs do not contain resource usage, so metric filters cannot produce cluster, service and task CPU and memory metrics.

  • d
    Turn on Container Insights with enhanced observability for the cluster in the ECS settings. Correct

    Container Insights is a cluster setting that works for Fargate and EC2 tasks and publishes cluster-, service- and task-level CPU, memory, network and storage metrics, plus the performance log events they are built from, to CloudWatch with no agent to run. Application logs still use a log driver such as awslogs.

The concept

Container Insights collects, aggregates and summarizes metrics and logs from containerized applications on Amazon ECS (including Fargate) and Amazon EKS. On ECS it is switched on per cluster or as an account default; the enhanced observability option adds task- and container-level detail.

Why that’s the answer

The constraints are Fargate, no collection code and least effort. Container Insights is a cluster setting that needs nothing added to Fargate tasks. A sidecar agent works but means editing and running extra containers in every task, Managed Service for Prometheus needs exporters and keeps the data outside CloudWatch, and log metric filters cannot measure resource usage. Container Insights does not ship application logs; those still need the awslogs or FireLens log driver.

How to reason it out
  1. Note the launch type: Fargate, so there are no container instances to install an agent on.
  2. Note the required granularity: cluster, service and task metrics in CloudWatch.
  3. Eliminate options that need code, exporters or extra containers in every task.
  4. Choose the cluster-level Container Insights setting.

Exam tip: ECS or EKS metrics at cluster/service/task level in CloudWatch with no agent to manage: turn on Container Insights.

CloudWatch Metrics, Alarms, and Log Filters for SOA-C03 — the lesson that teaches this.

Question 5Monitoring, Logging, Analysis, Remediation, and Performance Optimization

An application writes INFO lines continuously and occasional FATAL lines to a log file that the CloudWatch agent ships to a log group. The on-call team must be paged when five or more FATAL lines occur within 5 minutes, and the alarm must show OK, not INSUFFICIENT_DATA, during the many 5-minute periods with no FATAL lines. The alarm keeps the default missing-data treatment. Which configuration meets these requirements?

Choose one.

  • a
    Metric filter for FATAL with value 1 and no default value; alarm when the Sum is 5 or more over 5 minutes.

    Sum counts the FATAL lines correctly, but with no default value the filter publishes nothing in quiet periods, so under the default missing-data treatment the alarm drifts to INSUFFICIENT_DATA instead of OK.

  • b
    Metric filter for FATAL with value 1 and a default value of 0; alarm when the Average is 5 or more over 5 minutes.

    The default value of 0 keeps datapoints flowing, but the Average of a mix of 1s and 0s can never exceed 1, so the alarm never reaches 5 and nobody is paged.

  • c
    Metric filter for FATAL with value 1 and no default value; alarm when the Average is 5 or more over 5 minutes.

    This fails twice: the Average of values that are all 1 is 1, so the threshold is never met, and with no default value quiet periods publish nothing and the alarm goes to INSUFFICIENT_DATA.

  • d
    Metric filter for FATAL with value 1 and a default value of 0; alarm when the Sum is 5 or more over 5 minutes. Correct

    Each FATAL line adds 1, so the Sum over 5 minutes is the count of FATAL lines; the default value of 0 is published when ingested lines don't match, so quiet periods evaluate to 0 and the alarm reads OK.

The concept

A metric filter turns matching log events into a custom metric (value 1 per match). Two settings decide how an alarm on it behaves: the alarm statistic (Sum counts matches; Average of 1s is always 1) and the filter's default value (0 is published when log events arrive but none match; with no default value, quiet periods publish nothing).

Why that’s the answer

Two facts combine. Counting needs Sum, because every matching datapoint is 1 and Average never rises above 1. Reading OK in quiet periods needs datapoints in those periods: with the default missing-data treatment an alarm with no datapoints goes to INSUFFICIENT_DATA, so the filter needs a default value of 0. Only Sum with a default value of 0 satisfies both. Sum without a default value is the tempting near-twin: it pages correctly but sits in INSUFFICIENT_DATA most of the time.

How to reason it out
  1. Work out what each datapoint holds: the value 1 for each FATAL match.
  2. Pick the statistic that turns those datapoints into a count over the period: Sum.
  3. Ask what happens in a period with no matches: without a default value nothing is published.
  4. Recall that missing datapoints under the default treatment lead to INSUFFICIENT_DATA, so set a default value of 0.
  5. Alarm on Sum of 5 or more over 5 minutes with an SNS action.

Exam tip: Count log matches with Sum, and give the metric filter a default value of 0 so quiet periods read OK instead of INSUFFICIENT_DATA.

CloudWatch Metrics, Alarms, and Log Filters for SOA-C03 — the lesson that teaches this.

Question 6Monitoring, Logging, Analysis, Remediation, and Performance Optimization

After an incident, an engineer adds a metric filter that counts a specific error string in a log group and expects its metric to show yesterday's error spike. New errors since the filter was added appear as datapoints, but the metric has nothing for yesterday, although yesterday's log events are still visible in the log group. What explains this?

Choose one.

  • a
    The filter has no default value configured, so periods before it was set up are left empty.

    A default value only makes the filter publish 0 for periods with no matching events once it is running; it does not cause or prevent evaluation of older events.

  • b
    The filter was attached to one log stream, so it counts events only from the newest stream.

    Metric filters are defined on a log group and apply to all of its streams; the stream that wrote yesterday's events is not the reason.

  • c
    The filter's metric keeps 1-minute data for only a few hours before it rolls up to hourly.

    CloudWatch keeps 1-minute datapoints for 15 days, so yesterday's data would still be visible if it had ever been published.

  • d
    The filter only evaluates events ingested after it was created, so earlier events never count. Correct

    Metric filters are not retroactive: they process log events as they are ingested after the filter exists, so events from before the filter was created never produce datapoints.

The concept

Metric filters evaluate log events as they arrive. They do not backfill: events ingested before the filter was created never become datapoints, even if the events are still stored in the log group.

Why that’s the answer

The stem rules out retention (the events are still there) and a broken pattern (new matches are counted). What remains is the timing rule: the filter only sees events ingested after it existed. To count yesterday's errors, run a Logs Insights query over that time range instead.

How to reason it out
  1. Note the evidence: old events still exist and new matches are counted.
  2. Eliminate pattern and retention problems, which the evidence contradicts.
  3. Recall that metric filters are not retroactive.
  4. Use Logs Insights for historical counts; create filters before you need their metric.

Exam tip: Metric filters only count events ingested after they are created; use Logs Insights to analyse the past.

CloudWatch Metrics, Alarms, and Log Filters for SOA-C03 — the lesson that teaches this.

What SOA-C03 domain 1 tests, topic by topic

The official exam guide breaks Monitoring, Logging, Analysis, Remediation, and Performance Optimization into 3 topics. The question bank follows the same split, so a weak topic shows up as a cluster of misses you can go back and read.

Published SOA-C03 practice questions per topic in Monitoring, Logging, Analysis, Remediation, and Performance Optimization
TopicWhat it coversQuestions
Implement metrics, alarms, and filters by using AWS monitoring and logging servicesExam guide task 1.1. Monitoring and logging for serverless, compute, and AI workloads with CloudWatch, CloudTrail, and Amazon Managed Service for Prometheus; configuring and managing the CloudWatch agent for EC2, ECS, and EKS metrics/logs; CloudWatch alarms (including composite alarms) that invoke services directly or via EventBridge, with SNS notifications; shareable cross-account, cross-Region dashboards.38
Identify and remediate issues by using monitoring and availability metricsExam guide task 1.2. Analyzing performance metrics and automating remediation with CloudWatch, Lambda, Systems Manager, CloudTrail, Kiro, and AWS DevOps Agent; routing, enriching, and delivering events with EventBridge and troubleshooting event bus rules; creating and running custom and predefined Systems Manager Automation runbooks.38
Implement performance optimization strategies for compute, storage, and database resourcesExam guide task 1.3. Optimizing compute via performance metrics, resource tags, and AWS tools (EC2 instance selection, placement groups); analyzing and optimizing EBS performance and volume types; S3 performance strategies (DataSync, Transfer Acceleration, multipart uploads, lifecycle policies); selecting and optimizing shared storage (EFS, FSx, Amazon S3 Files, EFS lifecycle policies); monitoring RDS with Performance Insights and CloudWatch alarms and tuning with proactive recommendations and RDS Proxy.38
Total114

Revise Monitoring, Logging, Analysis, Remediation, and Performance Optimization before you drill it

Other SOA-C03 domains

Monitoring, Logging, Analysis, Remediation, and Performance Optimization: your questions

Monitoring, Logging, Analysis, Remediation, and Performance Optimization is domain 1 of the SOA-C03 exam guide and carries 22% of the scored content — the heaviest of the 5 domains. On a 65-question paper that works out to roughly 14 questions, though AWS does not publish an exact per-domain count and individual exam forms vary.

Source

The domain weight and topic list on this page come from the official SOA-C03 exam guide.