Observability & Monitoring

OpenTelemetry setup in Node.js uses the NodeSDK with auto-instrumentations and an OTLP trace exporter pointing to your collector endpoint. Load the tracing module first before other imports so instrumentation hooks apply correctly.

Logs record discrete events for debugging and auditing, metrics aggregate measurements for alerting and dashboards, and traces show request flow across services. OpenTelemetry unifies all three pillars in one instrumentation framework.

Correlation IDs propagate via middleware that reads or generates an x-trace-id header and includes it in every log entry. OpenTelemetry context propagation handles this automatically by extracting and injecting trace context into request headers.

Use the RED method for services: request rate, error rate, and duration. For infrastructure resources, apply the USE method: utilization, saturation, and errors, covering the four SRE golden signals of latency, traffic, errors, and saturation.

Alert fatigue happens when teams receive too many non-actionable alerts without runbooks or owners. Avoid it by using severity tiers, alerting on trends rather than static thresholds, and attaching runbook and dashboard links to every alert.

Yes, Azure Application Insights integrates through the @azure/monitor-opentelemetry package using a connection string. Equivalent integrations exist for AWS CloudWatch with X-Ray and GCP Cloud Monitoring with Cloud Trace.