Observability metrics
Metrics across runtime, builds, and databases.
Last updated: 2026-03-17
Forgeon collects metrics for:
- Runtime health (CPU, memory, requests)
- Build performance (durations, failures)
- Database usage (IOPS, storage, egress)
Where to see metrics in UI
- Project → Metrics
- Database → Cluster → Metrics
Core signals you should monitor first
| Signal | Healthy baseline | Warning threshold | | --- | --- | --- | | HTTP 5xx rate | Near 0% | > 1% sustained | | p95 latency | Stable by endpoint | Sudden +50% spike | | CPU usage | < 70% average | > 85% sustained | | Memory usage | Headroom > 20% | Frequent near-limit | | Build duration | Within normal range | 2x baseline growth |
How to read spikes correctly
- Correlate latency spike with deploy timestamp.
- Check whether error rate rose at the same minute.
- Compare CPU/memory and request volume.
- Open runtime logs for the exact time window.
- If only one instance is affected, inspect replica health.
Alerting recommendations
- Trigger alert if 5xx > 1% for 5 minutes.
- Trigger alert if p95 > baseline x 2 for 10 minutes.
- Trigger alert if no metrics arrive after active traffic.
- Trigger alert on repeated restart loops.
Troubleshooting
Metrics not showing
- Wait for 1–2 samples after first deploy
- Ensure traffic is hitting the runtime
Latency high, errors low
- Check downstream services (DB, third-party API, DNS).
- Check queue/backpressure before scaling blindly.
CPU low, errors high
- Usually app-level bug, bad env var, or dependency failure.
- Inspect runtime logs by request ID / timestamp.