Observability Patterns¶

Each microservice demonstrates a different observability instrumentation strategy, giving workshop participants hands-on experience with the full spectrum of AWS observability tools.

Distributed Tracing¶

Log Routing¶

• FireLens (Fluent Bit) sidecar on ECS tasks routes logs to OpenSearch (optional, requires ENABLE_OPENSEARCH=true)
• Container Insights on ECS and EKS clusters for infrastructure metrics
• OpenSearch ingestion pipeline for centralized log analytics

Metrics¶

• CloudWatch metrics from all services
• Prometheus metrics exposed by petfood-rs and petlistadoptions-py
• Application Signals SLOs on petsearch-java

Instrumentation by Service¶

payforadoption-go¶

Uses the OpenTelemetry Go SDK with a CloudWatch agent sidecar in OTLP mode. Traces are exported via OTLP to the agent, which forwards them to X-Ray. This demonstrates manual SDK instrumentation in Go with custom SQL span processing for Aurora correlation.

petlistadoptions-py¶

Uses ADOT auto-instrumentation via a CloudWatch agent sidecar. The Python/FastAPI application is instrumented without code changes, though manual configuration is required for FastAPI framework support.

petsearch-java¶

Uses the ApplicationSignalsIntegration L2 CDK construct for Application Signals. Combines auto-instrumentation with manual span creation and includes SLO definitions for monitoring service-level objectives.

petsite-net¶

Deployed on EKS with a CloudWatch agent providing Application Signals. The .NET application runs behind CloudFront with both regional and global WAF protection.

petfood-rs¶

Uses the OpenTelemetry Rust SDK with custom Prometheus metrics. Demonstrates manual instrumentation in Rust with both trace and metric signal types.

petfoodagent-strands-py¶

AI agent deployed on Bedrock AgentCore. Uses the Strands Agents SDK for natural-language food recommendations with built-in observability.