Alerts log
What a row tells you
Section titled “What a row tells you”Dashboard → Alerting → Alerts lists every alert the engine raised for your organisation, one row per event:
- When and which monitor, with its type.
- Type — Down, Partial outage (with the failing regions out of the total), Degraded, Recovered (with the total downtime), SSL expiry.
- Detail — HTTP status, latency, region.
- Notified via — one icon per channel that was called, ✓ delivered or ✗ failed. A dashed “no channel” badge means the alert was detected but nothing was attached to that monitor: nothing was sent, and the row is your reminder to fix it.
- Acknowledge — marks an alert as seen. It clears the sidebar badge and stops the next escalation level. “Acknowledge all” applies to the rows matching the current filters.
Filters: period (24 h, 7 d, 30 d, or the whole retention), unacknowledged only, type, “without channel”, monitor. Export CSV downloads the filtered list.
Clicking a row opens the detail drawer: detection → notifications → acknowledgement → linked incident, plus Create an incident (title, severity and monitor pre-filled, the alert is linked to the incident once created) and a shortcut to the monitor.
Alerts are kept for the retention of your plan — 90 days (Free), 180 (Pro), 365 (Max), 730 (Enterprise) — then archived automatically.
Escalation entries
Section titled “Escalation entries”When an escalation policy fires a later level, the log shows it on the original alert as a Level 2 / Level 3 badge (and Reminder for repeats) rather than as a new row: one outage, one row, its whole notification history inside.
Reading the log as a checklist
Section titled “Reading the log as a checklist”- Rows with the dashed “no channel” badge — monitors that alerted but had nothing attached. Fix them from the monitor’s page, or from Alerting → Channels → apply to monitors.
- ✗ next to a channel — the provider rejected the call (expired webhook, wrong token). Open the channel and use Send test.
- Unacknowledged rows piling up — nobody is on it, or your team doesn’t use acknowledge. Either is worth knowing.
- Recovered rows with a long downtime — candidates for a public incident post-mortem.