Skip to main content
Observer reads your metrics, decides when something is wrong, runs your status pages and keeps the record of each incident. It does not run on-call schedules, escalation policies or phone calls. Keep the tool that pages your team: Observer sends it every alert and closes the alert when the problem recovers.

Pick a destination

Pushover and ntfy are experimental channels. Every destination other than email is available from the Starter plan.

What stays with Observer

  • Deciding: a metric turns unhealthy only after its threshold holds for the confirmation time you set, so a single bad sample does not page anyone. See Thresholds and confirmation time.
  • Closing: when the metric recovers, the alert closes in your pager. A close is never blocked by filters or snooze. See How alerts open and close.
  • Telling customers: the status page and its subscribers get the incident, which you can scope to the customers it affected.
  • The record: what was measured during the incident stays with it, for SLA reports and reviews.

Choose what pages

Each destination chooses its own alerts. Under Alerts, the Paging preset sends what should wake someone, and Custom lets you choose the alerts yourself. Under Routing, narrow a destination by services, metrics and minimum severity. A team can send critical alerts to its pager and everything else to a chat channel.

Moving off an on-call tool

If the tool you page with is being retired, move paging first, then connect Observer to the new tool. For Grafana OnCall OSS, see Move off Grafana OnCall OSS.