APM Dashboards
API latency, error rate, and throughput per route.
Last updated 05/24/26
What this surface does
APM Dashboards chart API route latency (p50/p95/p99), error rate, and request volume sourced from `api_latency_samples` (migration 173). Each chart is per-route with time-range and percentile selectors.
When to use it
- Investigating a "site feels slow" report — find the route with the regression.
- Capacity planning — which routes are trending up in volume.
- Pre/post deploy comparison — did the new release introduce p95 regression on any hot path.
Key gotchas
- Samples are written by the API middleware. Routes that bypass middleware (static assets, edge functions) are not in the data.
- The default percentile is p95. p99 includes tail outliers (cold starts, slow DB queries) and is the right percentile for SLO alerts.
- Data retention is 30 days. Older data is archived — use the Backfill Launcher to replay if needed.
Related
- Sentry alerts runbook
- Alert Routing, Status Console.