Signal and tool selection
Observability uses measurements, records, and request traces to understand how a workload behaves. Choose a tool by the question and where the data lives.
How it works
- Use metricsA metric tracks numerical performance data over time., numerical measurements over time, to identify changes and set CloudWatch alarms. Use logs, records of events, for details.
- Query CloudWatch log groupsA log group contains log streams with shared retention, monitoring, and access settings., collections of streams with shared retention and access settings, with Logs Insights or use Athena SQL for data stored in S3.
- Use a traceA trace records a request and its calls to downstream components., a record of a request and its downstream calls, to investigate latency with X-Ray. Application Signals combines application metrics, correlated traces, and dependency maps. A service level objectiveA service level objective tracks a performance indicator against a chosen goal and threshold. (SLO) is a measurable performance target.
- For security and configuration, combine activity records, resource history, network traffic, and findings rather than expecting one service to answer every question.
| Question | Tool |
|---|---|
| Who made an AWS action? | CloudTrail |
| How was a recorded resource configured? Is it compliant? | AWS Config |
| What IP traffic reached a network interface? | VPC Flow Logs |
| Does a supported stack resource differ from its explicitly set template properties? | CloudFormation driftDrift is a difference between the actual and expected configuration of a stack resource. detection |
| Where are threats, vulnerabilities, or consolidated findings? | GuardDuty, Inspector, or Security Hub CSPM, respectively |
| Where can I query container metrics in PromQLPromQL is the Prometheus query language used to filter and aggregate metrics., the Prometheus query language? | Amazon Managed Service for Prometheus |
| Where can I see container metrics and logs? | CloudWatch Container Insights |
| Where can I visualize operational signals from multiple sources? | Amazon Managed Grafana |
| Where can I visualize business data in S3? | AWS Quick Sight |
| Where can independent applications consume streaming data? | Kinesis Data Streams |
| Where can I buffer and deliver streaming data to S3 or OpenSearch? | Amazon Data Firehose |
| Where can I analyze and visualize logs with a search engine? | Amazon OpenSearch Service |
| Where can I process large datasets with Spark or Hadoop? | Amazon EMR |
| Where can I discover sensitive data in S3? | Amazon Macie |
| Where can I investigate suspicious security findings? | Amazon Detective |
| Where can I find resources shared with external entities? | IAM Access Analyzer |
| Where can I control web requests to protected applications? | AWS WAF |
Exam cues
- If the question says query archived S3 data with SQL, think Athena, because it queries data directly in S3.
- If the question says who changed a security group and what it looked like before, think CloudTrail and AWS Config, because they provide action records and configuration history.
- If the question says simulate a user journey without traffic, think Synthetics, because canaries run scheduled scripts.
- If the question says actual user sessions, think RUM, because it collects client-side session data.
- If the question says export CloudWatch metrics continuously to S3, think metric streams, because they deliver metrics in near real time.
- If the question says observe several linked accounts centrally, think CloudWatch cross-account observability, because the monitoring account can view their metrics, logs, and traces.
Remember
- GuardDuty detects threats; Inspector scans workloads for vulnerabilities and unintended exposure; Security Hub CSPM aggregates findings and checks security controls.
- Drift detection compares supported resource properties that are explicitly set, not every possible property.
Trap: CloudTrail identifies actions and actors, but AWS Config supplies the recorded resource configuration history. Neither replaces VPC Flow Logs for traffic information.
Da documentação da AWS (31)
- CloudWatch metrics track performance data, and alarms monitor metrics against thresholds.
- CloudWatch Logs Insights interactively searches and analyzes CloudWatch Logs data.
- A CloudWatch log group contains log streams with shared retention, monitoring, and access settings.
- Athena queries data directly in S3 using SQL.
- AWS Quick Sight provides business intelligence and data visualization.
- AWS Quick Sight supports S3 data sources.
- Amazon Managed Grafana queries, correlates, and visualizes metrics, logs, and traces from multiple sources.
- CloudTrail records actions by users, roles, and AWS services and can identify the actor and time.
- AWS Config records supported resource configurations and history and evaluates rules for compliance.
- VPC Flow Logs capture information about IP traffic to and from VPC network interfaces.
- CloudFormation drift detection compares actual supported resource properties with expected explicitly set properties.
- Amazon Managed Service for Prometheus stores container metrics and supports PromQL queries.
- CloudWatch Container Insights collects and summarizes metrics and logs for containerized applications.
- X-Ray uses traces of requests and downstream calls to identify latency and bottlenecks.
- Application Signals provides application metrics, correlated traces, dependency maps, and service level objectives.
- A service level objective tracks a performance indicator against a chosen goal and threshold.
- Synthetics canaries are scheduled scripts that simulate customer actions.
- CloudWatch RUM collects client-side performance data from actual user sessions.
- GuardDuty detects threats by analyzing AWS data sources and logs.
- Inspector scans EC2 instances, ECR images, and Lambda functions for software vulnerabilities and unintended network exposure.
- Security Hub CSPM aggregates findings and runs checks against security controls.
- Kinesis Data Streams supports multiple applications consuming a stream independently.
- Amazon Data Firehose buffers streaming data and delivers it to destinations including S3 and OpenSearch Service.
- Amazon OpenSearch Service supports log analytics and visualization with OpenSearch Dashboards.
- Amazon EMR runs big data frameworks including Apache Hadoop and Spark for data processing.
- CloudWatch metric streams deliver metrics continuously to destinations including S3, with JSON and OpenTelemetry formats available.
- Amazon Macie discovers sensitive data in S3 and monitors bucket security and access control.
- Amazon Detective helps investigate security findings using aggregated data and visualizations.
- IAM Access Analyzer can identify resources shared with external entities.
- AWS WAF monitors HTTP and HTTPS requests and controls access to protected web resources.
- CloudWatch cross-account observability lets a central monitoring account view metrics, logs, and traces from linked source accounts.
Incident correlation and bottlenecks
Incident correlation compares evidence from the same period to narrow a cause. A bottleneck is a component that constrains the progress of a request or pipeline.
How it works
- Find the affected time window in application metricsA metric tracks numerical performance data over time. and search logs for errors.
- Inspect correlated traces and the application dependency map. A traceA trace records a request and its calls to downstream components. follows a request through its components; the X-Ray trace map can reveal slow downstream calls.
- Compare AWS action times in CloudTrail with the incident, then investigate changes and validate fixes with logs. A timing match is a lead, not proof of cause.
- For stopped or failed workloads, inspect the service’s own activity or task details before choosing a fix.
Exam cues
- If the question says a canaryA canary is a scheduled script that monitors endpoints by following customer actions. fails after a release, think failed run logs, screenshots, and step reports, because they show the failing steps when available.
- If the question says an Auto Scaling launch failed, think activity history, because it reports the scaling activity status.
- If the question says an ECS task stopped, think stopped task details and service events, because they record stop reasons and service actions.
- If the question says an SQS to Lambda to DynamoDB pipeline is delayed, think queue age, Lambda throttles and concurrency, and DynamoDB throttled requests, because they expose pressure at different stages.
Remember
- SQS
ApproximateAgeOfOldestMessageis an approximate queue age. - Lambda
IteratorAgeapplies to DynamoDB, Kinesis, and DocumentDB event sources, not SQS. - Lambda
Throttlescounts throttled requests;ConcurrentExecutionscounts function instances processing events.
Trap: Do not use Lambda
IteratorAgeas the SQS backlog age. Use the SQS age metric for that queue.
Check queue age and Lambda invocation pressure.
Check DynamoDB throttled requests.
Da documentação da AWS (13)
- Application Signals provides application metrics, correlated trace spans, and a map of applications and dependencies.
- The X-Ray trace map can reveal bottlenecks and latency spikes in downstream calls.
- CloudWatch Logs Insights can help identify potential causes and validate fixes.
- CloudTrail records AWS actions and provides actor and time information.
- A failed Synthetics canary run can show logs, screenshots, and step reports when available.
- Auto Scaling activity history shows the status of scaling activities including success or failure.
- ECS stopped task details provide the stop code and reason, while service events record service actions.
- Container Insights provides diagnostic information including container restart failures.
- SQS ApproximateAgeOfOldestMessage measures the age of the oldest unprocessed message in a queue.
- Lambda Throttles counts throttled invocation requests.
- Lambda ConcurrentExecutions counts function instances processing events.
- Lambda IteratorAge measures record age for DynamoDB, Kinesis, and DocumentDB event sources.
- DynamoDB ThrottledRequests measures requests that were throttled.
Flashcards
Nada para revisar aqui agora. Escolha outros tópicos ou outro filtro.
·
Todos os 17 flashcards em lista
A team needs business analysis of data in S3 and a dashboard combining monitoring sources. Which services fit?
Use AWS Quick Sight for business visualization and Amazon Managed Grafana for operational dashboards across sources.
A team needs to analyze CloudWatch log groups during an incident and query archived S3 data with SQL. Which tools fit?
Use CloudWatch Logs Insights for log groups and Athena for SQL over S3 data.
Which AWS services show who changed a security group and its earlier configuration?
CloudTrail records the actor and action; AWS Config provides recorded configuration history.
Which AWS tools show a security group change actor, configuration history, network traffic, and stack drift?
CloudTrail, AWS Config, VPC Flow Logs, and CloudFormation drift detection, respectively.
Where can a team query Kubernetes container metrics in PromQL, inspect container logs, and investigate distributed traces on AWS?
Use Amazon Managed Service for Prometheus for PromQL metrics, Container Insights for logs and metrics, and X-Ray for traces.
How do GuardDuty, Inspector, and Security Hub CSPM differ?
GuardDuty detects threats, Inspector scans workloads for vulnerabilities and unintended exposure, and Security Hub CSPM aggregates findings and checks security controls.
A team needs to monitor a checkout journey even when nobody is using the application. What AWS feature fits?
CloudWatch Synthetics canaries run scheduled scripts that simulate customer actions.
A team needs to investigate actual browser session performance rather than simulated checks. What AWS feature fits?
CloudWatch RUM collects client-side data from real user sessions.
A team needs to deliver streaming logs to S3 without managing a delivery application. Which AWS service fits?
Amazon Data Firehose buffers and delivers streaming data to S3.
A security team needs to discover sensitive data in S3 and investigate suspicious findings. Which AWS services fit?
Amazon Macie discovers sensitive S3 data; Amazon Detective helps investigate security findings.
How can a team continually export CloudWatch metrics to S3?
Use a CloudWatch metric stream, with JSON or OpenTelemetry output.
An application error rate rises after a release. How can an AWS team investigate?
Compare application metrics, logs, correlated traces, and CloudTrail actions over the affected time. Use the trace map to inspect slow dependencies.
A CloudWatch Synthetics login canary starts failing after a release. What should the team inspect?
Inspect the failed run's logs, screenshots, and step reports when available, and compare them with a successful run.
An Auto Scaling group cannot add instances and an ECS task stops. Where should the team look?
Check Auto Scaling activity history and the ECS stopped task's stop code, reason, and service events.
An SQS to Lambda to DynamoDB pipeline is delayed. Which metrics help locate the bottleneck?
Check SQS
ApproximateAgeOfOldestMessage, LambdaThrottlesandConcurrentExecutions, and DynamoDBThrottledRequests. LambdaIteratorAgeis for supported stream sources, not SQS.Which AWS tool can show whether a downstream dependency is causing a latency spike?
The X-Ray trace map shows downstream calls and helps identify latency spikes and bottlenecks.
A team suspects repeated container restarts in an EKS workload. Where can it see diagnostic information?
CloudWatch Container Insights provides diagnostic information including container restart failures.