Health Monitor

Last updated May 27, 2026

The same surface as Server Status but exposed at machine-readable endpoints for external monitoring.

Endpoints

Endpoint	Auth	Returns
`GET /api/v1/health`	none	`{ "status": "ok" \| "degraded" \| "down", "version": "...", "uptime_s": N }`, used as the basic liveness probe.
`GET /api/v1/health/detailed`	admin Bearer / cookie	Per-service status (same as `/jabali-admin/server-status`) as JSON.
`GET /metrics`	admin Bearer	Prometheus-format metrics (request counts, latencies, reconciler tick durations, queue depths).

Status semantics

status	Meaning
`ok`	All watched services healthy.
`degraded`	A non-critical service is failed/degraded (e.g. ClamAV freshclam stale > 7 d). Panel still serves.
`down`	A critical service is failed (panel-api itself returning health, so MariaDB / nginx / Stalwart down counts here).

Watched services

jabali-panel.service
jabali-agent.service
nginx.service
mariadb.service
postgresql.service (only if at least one user has a Postgres DB; otherwise ignored)
pdns.service
pdns-recursor.service
stalwart-mail.service
kratos.service
bulwark.service
redis.service
crowdsec.service

The watched set is computed at startup; services that aren’t installed don’t count against degraded.

Use with an external monitor

UptimeRobot / Pingdom / a self-hosted Uptime-Kuma:

Point at https://<panel-hostname>/api/v1/health, anonymous, fast.
Set expected response: HTTP 200 + body contains "status":"ok".

For Prometheus scraping, point at /metrics with a bearer token (mint under /jabali-admin/automation).

Notifications integration

The service_down event source (M14) reads the same internal state, so you don’t need an external monitor for in-house alerting. The external monitor is useful for “the panel-api itself is down, who tells me?” cases.