# Portfolio Observability Incident Runbook

## Target unavailable
1. Confirm DNS and TLS from a second network.
2. Check the latest Cloudflare Pages deployment and origin status.
3. Correlate the target-down alert with Loki logs and the Tempo probe trace.
4. Roll back the last release when the failure started after deployment.
5. Resolve the incident only after consecutive successful checks.

## High latency
1. Compare p50, p95, and p99 latency.
2. Inspect slow traces and identify the longest span.
3. Search logs using the trace ID.
4. Check saturation, upstream APIs, and database operations.
5. Document remediation and validate the SLA budget.

## Telemetry pipeline silent
1. Validate OTLP endpoint reachability.
2. Check collector queue, exporter errors, and credentials.
3. Verify Prometheus targets and Loki labels.
4. Confirm Tempo is accepting spans and object storage is healthy.