Monitors

Creating a monitor

Fill in Name, Kind, Target, Interval (≥ 30 s), and Timeout. The background scheduler (meridian-beat) picks it up within one beat tick (∼ 60 s) and starts sampling.

Kinds

• https / http — HEAD by default, 2xx/3xx = OK, timeout/5xx = down, cert errors trigger cert-specific notifications.
• port_tcp — bare TCP connect. Use for things like SMB/SSH/databases where you just want to know the port answers.
• ping_icmp — ICMP echo. Requires the host to respond; many cloud providers block it.
• cert_expiry — pulls the cert and alerts at 30/14/7-day thresholds.
• dns_record — checks that a record resolves to a specific expected value.

Editing + deleting

Each row has Edit (inline form pre-populated with current values), Disable/Enable, and Delete. Deleting a monitor removes its samples and incident history too.

Notifications

Monitors dispatch monitor.incident.opened on state change to bad, and monitor.incident.closed on recovery. Attach notification channels in Admin → Integrations → Notifications.

Gotchas

• HTTPS against an IP (not a hostname) will fail TLS hostname verification. Use the hostname or set kind = port_tcp.
• Interval < 30 s is rejected — beat runs per minute, so anything tighter is wishful thinking.
• Consecutive-fails is what triggers an alert, not a single sample — defaults to 2.