On-call rotations and escalation policies that actually page

On-call rotations and escalation that actually page

Setup

1. Teams first. Create the team, add members, define the rotation: who is on, handoff schedule, time zone handling. A rotation with no backup is a single point of failure with extra steps.
2. Escalation policy: step 1 pages the on-call, step 2 (after N minutes unacked) pages the secondary or the manager, step 3 the wider team. Each step gets a timeout. No step means the page dies silently with the person who is camping.
3. Wire monitors to the policy, not to individuals. People join and leave; the policy persists.
4. Incident integration: incidents can page teams directly from the incident, and pages can escalate into incidents. Decide the direction per severity: SEV-1 auto-creates the incident, lower sevs page first.

Discipline

• Ack means owned. An acked page that nobody works is worse than an unacked one (it stopped escalating). The ack contract: ack, then update within 15 minutes.
• Rotation hygiene: no one on-call two weeks running, no rotation with one member. Review the schedule monthly, not when someone burns out.
• Page audit quarterly: pull the page history, kill the monitors that page for non-actionable conditions, and tune the ones that page weekly for the same known issue.
• Test the chain after every rotation change: page the schedule in staging and confirm the right phone buzzes.

Verify

A test page from a monitor reaches the on-call device within the step-1 timeout, escalates when unacked, and the incident it creates carries the monitor context.

The Datadog On-Call docs cover rotations, schedules, and escalation policies; Incident Management integrates with On-Call to escalate pages into incidents and page teams from an incident. Verified against live docs 2026-09-26.