Monitoring

Last updated

Three signal sources, complementary.

1. Audit log

/jabali-admin/audit. Append-only structured rows. Every privileged mutation lands here:

  • Subject (user being acted on).
  • Actor (operator / user / system that initiated).
  • Source (UI / CLI / reconciler).
  • Action (domain.create, db.root.rotate, mailbox.passwd, etc.).
  • Target (resource id).
  • Result (ok / fail + structured error code).

Use for: forensic “who did what when” investigation; compliance.

CLI:

jabali audit query                     # the last 50 events
jabali audit query --limit 500         # up to 1000
jabali audit query --q db.             # search action, target, actor kind and result
jabali audit verify                    # check the hash chain for tampering

2. Notifications

/jabali-admin/notifications. Event sources fan out to the configured channels: email, Slack, Discord, Telegram, ntfy, webhook, SMS and Web Push (see notifications.md).

Use for: human-attention-needed alerts in real time.

3. Health and server status

The panel has no /metrics endpoint and no Prometheus exporter. An external monitor polls the endpoints in Health Monitor: GET /health and GET /health/agent (no sign-in), and the automation endpoints GET /api/v1/automation/status (scope read:status) and GET /api/v1/automation/server-status (scope read:metrics).

Use for: uptime checks and fleet dashboards.

Tail-the-logs alternative

For ad-hoc:

journalctl -u jabali-panel -u jabali-agent -u nginx -u jabali-stalwart -f

Structured JSON in stdout from panel-api and agent; jq away.

What’s not included

  • APM / tracing — not yet. OpenTelemetry support is on the roadmap; the panel + agent currently emit log lines but not trace spans.
  • Continuous profiling — not shipped.
  • eBPF observability — not shipped. The Tetragon tripwires were removed.