SAP-C02
Deep Dive
Observability is the ability to understand a system's internal state from its external outputs. On the SAP-C02, observability questions go beyond basic CloudWatch — they test complete monitoring architectures for distributed systems at scale.
Contents
Metrics
Numeric data aggregated over time. How much? How fast? How many errors?
Logs
Event records with full context. What exactly happened, and when?
Traces
Following requests across multiple services. Where's the latency?
Advanced metrics
Advanced CloudWatch Logs
Container Insights
Collects metrics and logs from containers on ECS, EKS, and Kubernetes. Aggregates data at the cluster, node, pod, and individual container level.
Lambda Insights
Advanced performance monitoring for Lambda functions. Diagnoses duration, throttling, and memory error issues.
X-Ray generates a visual map of your microservices architecture showing how requests flow between services, where the latency is, and what errors occur. Essential for debugging distributed systems.
Key X-Ray concepts
Integration with AWS services
Lambda
Native — enable Active Tracing on the function
API Gateway
Native — enable X-Ray Tracing on the stage
ECS/EC2
Requires the X-Ray daemon as a sidecar or daemon
EKS
ADOT Collector as a DaemonSet or sidecar
ECS Service Connect
The managed proxy publishes traffic metrics automatically
Elastic Beanstalk
Native — configure via .ebextensions
CloudWatch ServiceLens
ServiceLens combines X-Ray traces, CloudWatch metrics, and CloudWatch logs into a unified view. It generates a Service Map showing the health of each service with latency, error, and request-rate metrics — no need to jump between multiple consoles.
CloudWatch Evidently was retired on October 17, 2025. Today feature flags are managed with AWS AppConfig (Systems Manager): it lets you roll out features gradually, validate configuration, and roll back automatically if a CloudWatch alarm fires. A/B experiments are built by combining the flags with your own business metrics.
Feature Flags
AppConfig enables features for a % of users without redeploying code. Enables controlled canary releases.
A/B Experiments
AppConfig variants are combined with business metrics in CloudWatch to measure impact.
Automatic rollback
AppConfig reverts the configuration if a CloudWatch alarm fires during deployment.
Segmentation
Shows features based on user attributes: region, plan, app version.
| Event type | What it records | Cost | Use case |
|---|---|---|---|
| Management Events | Control plane operations: creating/modifying/deleting resources | Free (first trail) | Auditing infrastructure changes — "who created this resource" |
| Data Events | Data-plane operations: S3 GetObject, Lambda Invoke, DynamoDB GetItem | Paid | Auditing access to sensitive data, post-incident forensics |
| Insights Events | Unusual activity: spikes in IAM errors, unusual API calls | Paid | Detecting anomalous behavior in API calls |
CloudTrail Organization Trail
An Organization Trail created in the Management Account records events from every member account into a centralized S3 bucket. Member accounts cannot modify or disable the trail. This guarantees complete, immutable auditing for the whole organization.
Pitfall: "X-Ray records every request by default"
Reality: FALSE. X-Ray uses sampling — by default it records 5% of requests + 1 request/sec. The sampling rate can be increased for complete debugging.
Pitfall: "CloudWatch and CloudTrail are equivalent"
Reality: These are completely different. CloudWatch monitors performance (metrics, app logs). CloudTrail audits AWS API calls (who did what with which resources).
Pitfall: "Container Insights works with no extra configuration on EKS"
Reality: FALSE. On EKS it requires deploying the CloudWatch Agent (or Fluent Bit) as a DaemonSet. On ECS with Fargate it requires a sidecar container with the CloudWatch Agent.
Did you get this topic?
Test what you just learned
A company has a microservices architecture with 15 services on Lambda and ECS. Users report that some requests intermittently take more than 10 seconds. The operations team can't determine which specific service is causing the latency. Which AWS service directly solves this problem?
Put what you learned to the test with original SAP-C02 exam questions.
Start SAP-C02 mock examUnlimited mock exams with the Pro plan
Log in to track your progress.