Health and metrics
Endpoints
| Path | Purpose | Authentication |
|---|---|---|
| /q/health/live | Process liveness | Public |
| /q/health/ready | Runtime readiness | Public |
| /q/health | Aggregate health | Public |
| /q/metrics | Prometheus and runtime metrics | Public |
Public means the application does not require OIDC for these paths. Restrict them with cluster network policy, a private service, or your monitoring gateway.
Custom capacity gauges
DonkeyFleet currently publishes these byte-valued gauges:
| Metric |
|---|
| donkeyfleet.capacity.unprotected.provisioned |
| donkeyfleet.capacity.unprotected.physical |
| donkeyfleet.capacity.orphaned.provisioned |
| donkeyfleet.capacity.orphaned.physical |
The values are updated from complete observation. Provisioned and physical gauges answer different questions; alert and graph them separately.
Operational signals
Monitor more than HTTP readiness:
- age and status of the last reconcile;
- skipped_observe_incomplete, skipped_peer_down, and orphan guardrail outcomes;
- critical notifications and severity escalations;
- approval backlog and long-running ONTAP jobs;
- cluster reachability probe age;
- PostgreSQL, Vault, OIDC, and ONTAP endpoint health.
Detailed DonkeyFleet reconcile and action counters are not currently exported as custom metrics, so use the console, audit history, and structured logs alongside Prometheus.