p50 / p95 / p99
forever.
Response time percentiles, query volume analytics, SLO burn rates, PromQL export, and error budget tracking -- retained indefinitely.
Every endpoint.
Every percentile.
p50 tells you the typical experience. p95 tells you the slow experience. p99 tells you the tail. All three, for every endpoint, retained forever.
| Endpoint | p50 | p95 | p99 | RPM |
|---|---|---|---|---|
| /api/v1/zones | 12 | 38 | 142 | 4,200 |
| /api/v1/records | 8 | 22 | 89 | 12,800 |
| /api/v1/certs | 340 | 1200 | 3400 | 480 |
| /api/v1/monitors | 15 | 44 | 210 | 2,100 |
| /api/v1/incidents | 18 | 52 | 188 | 890 |
| /api/v1/status-pages | 6 | 14 | 42 | 18,400 |
Every query.
Counted.
See DNS query volume by hour, day, or month. Break down by record type, zone, or geographic origin. Export to CSV or query via API.
Burn rate above 1x?
Fix it now.
Define SLOs, and Relays continuously computes burn rates. A burn rate above 1.0x means you are consuming your error budget faster than planned -- time to act.
| SLO | Target | Current | Burn Rate | Status |
|---|---|---|---|---|
| API Availability | 99.95% | 99.98% | 0.4x | healthy |
| DNS Resolution p99 | <50ms | 38ms | 0.6x | healthy |
| Cert Issuance p95 | <10s | 4.1s | 0.2x | healthy |
| Status Page TTFB | <200ms | 142ms | 0.8x | elevated |
| Incident MTTA | <5m | 3m 12s | 1.2x | elevated |
Your Grafana.
Our metrics.
Relays exposes a Prometheus-compatible /metrics endpoint. Scrape it into your existing Grafana stack, or use our built-in dashboards.
How much risk
can you spend?
Error budgets quantify the gap between your SLO and perfection. Spend it on risky deploys. Save it for reliability. Relays tracks the balance in real time.