Skip to Content
Cloud Engineering Academy
11. Monitoring, Logging & Troubleshooting

Module 11 cheat sheet

One-page revision summary for Monitoring, Logging & Troubleshooting: services, Azure mapping, CLI commands, limits, decision rules, exam traps and cost traps.

SOA-C03DOP-C02DVA-C03

Assumes: AWS CLI v2 · AWS SDK for .NET v4 · OpenTelemetry .NET · .NET 10 LTS · Region us-east-1 · facts as stated in the Module 11 lessons; items marked VERIFY there are not repeated as certain here. Verified against official sources on 2026-11-18.

A printable summary of Module 11. Each row links back to the lesson with the detail.

Key services

ServiceWhat it is for
CloudWatch LogsLog groups with per-group retention and class (Standard or Essentials); metric filters and subscription filters
Logs InsightsAd hoc queries over log groups; billed by data scanned, so narrow the time range
CloudWatch metrics + EMFTechnical and custom business metrics; EMF turns a JSON log line into metrics with no API call
Alarms / composite alarmsMetric alarms with datapoints-to-alarm; composites combine alarms so one incident is one page
Dashboards + SyntheticsOne on-call screen; canaries probe from outside (Node.js, Python or Java runtimes, no .NET)
X-Ray + OpenTelemetry (ADOT)Distributed traces; instrument with OpenTelemetry, X-Ray is a backend
Application SignalsService map, standard metrics, SLOs and burn-rate alarms built from OpenTelemetry data
CloudTrailWho did what, when, from where; event history is 90 days of management events, free
SNS / EventBridgeAlarm actions and event-driven alerts (for example a risky API call)
Budgets / Cost Anomaly DetectionCatch CloudWatch cost mistakes within a day

.NET / Azure → AWS

Microsoft / AzureAWS
Application InsightsApplication Signals + X-Ray over OpenTelemetry
Log Analytics + KQLCloudWatch Logs + Logs Insights query language
TelemetryClient.GetMetricEMF line or PutMetricData
Azure Monitor alerts + action groupsAlarms + SNS (composite alarms to cut noise)
Availability testsSynthetics canaries
Workbooks / dashboardsCloudWatch dashboards (keep in Git)
Activity LogCloudTrail event history and trails
Basic Logs table planEssentials log class
ActivitySource / ActivityOpenTelemetry spans, W3C traceparent

Most useful CLI commands

TaskCommand
Tail a log groupaws logs tail /hl-m11/transfers-api --follow --since 10m
Set retentionaws logs put-retention-policy --log-group-name NAME --retention-in-days 7
Run a Logs Insights queryaws logs start-query --log-group-name NAME --start-time T --end-time T --query-string '...' then aws logs get-query-results --query-id ID
Publish a metricaws cloudwatch put-metric-data --namespace Harbourline/Payments --metric-name TransferCount ...
Create / combine alarmsaws cloudwatch put-metric-alarm ... · aws cloudwatch put-composite-alarm ...
Rehearse a pageaws cloudwatch set-alarm-state --alarm-name NAME --state-value ALARM --state-reason TEXT
Read alarm historyaws cloudwatch describe-alarm-history --alarm-name NAME --history-item-type StateUpdate
Who changed it?aws cloudtrail lookup-events --lookup-attributes AttributeKey=EventName,AttributeValue=NAME
List SLOsaws application-signals list-service-level-objectives
Anything left over?aws resourcegroupstaggingapi get-resources --tag-filters Key=Project,Values=cea-lab
Live .NET countersdotnet-counters monitor --process-id PID

Key limits and defaults

ItemValueLesson
Log group retention defaultNever expires; set it deliberatelyCloudWatch Logs for .NET
Log event size / PutLogEvents batch1,024 KB / 1 MBCloudWatch Logs for .NET
Metric filters / subscription filters per log group100 / 5CloudWatch Logs for .NET
StartQuery rate10 per secondCloudWatch Logs for .NET
Essentials log classNo metric filters, subscription filters or EMF; class fixed at creationLog aggregation and error monitoring
EMF eventSingle valid JSON object; up to 30 dimensions per metric (cost, not count, is the limit)CloudWatch metrics and custom metrics
Each dimension combinationA separately billed metricCloudWatch metrics and custom metrics
Custom metric price (US East)USD 0.30 per metric per month for the first 10,000CloudWatch metrics and custom metrics
Anomaly detectionAbout two weeks of history to learnAlarms and composite alarms
CloudTrail event history90 days, management events only, one Region at a timeCloudTrail for operations
X-Ray trace ID1-<8 hex time>-<24 hex>; a W3C trace ID is reformatted to itDistributed tracing

Decision rules

If…Then…
Deciding what to alert onCustomer-visible symptom, sustained over several periods, with a named action
Failures only publish a metric when they happennotBreaching for missing data
Traffic vanishing means troublebreaching (traffic floor), and alarm the canary the same way
Several alarms describe one incidentComposite alarm; put actions on the composite
A rate with small traffic is noisyAdd a minimum-traffic guard
Metric has a daily or weekly shapeAnomaly detection; otherwise a static threshold
A business count is neededEMF from the app, not a log scan
Need a request across servicesOpenTelemetry traces with W3C propagation; scrub attributes
Need an SLOSLI, goal, window; page on fast burn, ticket on slow burn
Log group feeds alarms or EMFStandard class, never Essentials
Logs needed across accountsRead in place with cross-account observability; copy only for retention or SIEM
Something changed with no deployCloudTrail first: who, which credentials, from where
Debugging an incidentSymptom, hypothesis, evidence, fix; mitigate before root cause
Needing an incident toolIncident Manager is closed to new customers (2025-11-07); alarms to SNS or third-party on-call

Common exam traps

  • Event history is not a trail. 90 days, management events only, per Region. Longer retention or data events need a trail.
  • CloudTrail is not an application log and not a business audit trail.
  • Log groups never expire by default, and the log class cannot be changed later.
  • Essentials (formerly Infrequent Access) loses metric filters, subscriptions and EMF.
  • EMF needs a pure JSON line; logging it through a normal logger that adds text breaks it.
  • Each dimension combination is its own metric; a user ID as a dimension is a cost and cardinality bug.
  • Missing data is a decision: missing leaves state unchanged, notBreaching and breaching force it.
  • Use Sum for counters, percentiles for latency, not Average.
  • Synthetics has no .NET runtime; a Node.js or Python script calls your endpoint.
  • ADOT .NET SDK does not support AWS SDK for .NET v4; check before relying on automatic spans.
  • Logs Insights is for investigation; use metrics and alarms for continuous checks.

Cost traps

TrapAvoid it by
Ingestion of chatty logs (Standard class, USD 0.50 per GB in US East)Log levels, sampling, retention, Essentials for rarely read logs
High-cardinality dimensionsPrice the series count before deploying
Debug logging left on in productionAlarm on IncomingBytes; also a data-protection problem
Logs Insights over 30 days "to be safe"Start at 15 to 60 minutes and filter early
Canaries left runningThey bill per run; delete after the lab
Collector-less X-Ray export at default 100% samplingSample low; put a collector in front
Extra CloudTrail trails and data eventsOne organisation trail; select data events narrowly
Cost surprises noticed at month endCloudWatch-scoped budget and anomaly detection

Independent learning resource — not affiliated with, endorsed or sponsored by Amazon Web Services. AWS and related marks are trademarks of Amazon.com, Inc. or its affiliates. All banking data in examples is fictional. Sample architectures are for learning and are not automatically compliant with GDPR, PCI DSS or any banking regulation.

On this page