A new module for endpoint health: Rate, Errors, and Duration metrics for every HTTP and RPC endpoint your services handle.
What each row shows:
- A protocol badge (HTTP, gRPC, or Connect)
- Throughput, error rate, and p50/p95/p99 latency
- For HTTP: a full 2xx/3xx/4xx/5xx status breakdown
One detail worth knowing:
4xx responses count as client faults, not endpoint errors, so bad requests from callers don't make a healthy endpoint look broken. Only 5xx responses and error-marked spans count toward the error rate.Click any endpoint for time-series charts and its recent (or failed-only) traces, with a shareable link.
This should give endpoint-level visibility into which of your endpoints are slow, failing, or busy, without building a synthetic check for every route.
Check out the API Monitoring documentation for full details.