Alert routes, on-call & maintenance
One alerting pipeline for everything that fires — monitors, hosts, watches, report rules. Match alerts to routes, escalate through tiers, page the person on call, and suppress noise during planned work.
- Read
- 6 min
- Sections
- 3
- Plans
- Routes Pro+
Alert routes
A route decides where an alert goes. Each route has a match (by severity — info, minor, major, critical — and source — monitor, host, watch, report rule) and an ordered list of escalation steps. Routes are evaluated by priority; the first matching route wins.
Each step targets a delivery channel (Slack, Discord, webhook, PagerDuty, Opsgenie, …) and a delay. Tier 1 fires immediately; later tiers fire after their delay if the alert is still unacknowledged — that’s your escalation.
POST /api/v1/alert-routes
Authorization: Bearer <token>
Content-Type: application/json
{
"name": "Critical → PagerDuty",
"priority": 0,
"matchJson": "{\"severity\":\"critical\",\"source\":\"monitor\"}",
"stepsJson": "[
{\"channelId\":\"slack-uuid\",\"delaySec\":0},
{\"channelId\":\"pagerduty-uuid\",\"delaySec\":300,\"onCallScheduleId\":\"sched-uuid\"}
]"
}On-call schedules
An on-call schedule is a rotation of workspace members with a timezone and a cadence (daily or weekly). The order you pick members in is the rotation order. An escalation step can target “whoever is on call” instead of a fixed person — DataRunner resolves the current on-call member at fire time, and one-off overrides cover swaps.
Maintenance windows
Planned work shouldn’t page anyone. A maintenance window suppresses alerts for a chosen scope — the whole workspace, a single monitor, or a single host — between a start and end time. Anything that would have fired inside the window is silenced; everything resumes automatically when it ends.