Build the rotation once and it keeps turning. Everyone can see whose week it is, overrides and swaps cover holidays and dentist appointments, the rota subscribes into your own calendar, and the escalation policy pages the person the roster names - not the whole channel.
primary
backup
Thu-Fri backup is an override - the underlying rotation is untouched and resumes on Saturday.
Incident fires -> on-call roster consulted -> escalation chain paged until ACKed -> incident closed.
Rotate daily or weekly, and set how many days or weeks one shift lasts - anything from a single day up to 52 weeks. Set the members, the order, and the handover time, and the rota keeps turning without anyone re-typing a spreadsheet. Everyone takes their fair share, the schedule is always visible, and your team knows exactly who's on call and when their turn comes around.
People take holidays. People get sick. People swap shifts because they have a dentist appointment on Tuesday. Overrides let you handle all of it without breaking the underlying rotation - a one-time override for a specific date range, after which the original schedule resumes exactly where it would have been. A shift swap is booked as a pair, so cancelling one half cancels both and nobody is left covering a shift they never agreed to.
Overrides are an overlay rather than an edit, so everything that asks who is on call reads the same answer: the escalation policy, the on-call indicator, the timeline, and the calendar feed.
Platform, database, and application on-call rarely belong to the same people. Create as many rosters as you need, each with its own members and cadence, and let escalation policies pick the one that fits the alert. Handover times are timezone-aware, so a distributed team hands over at the right local hour.
If the on-call engineer doesn't respond in five minutes, escalate to the next person. Then the next. Different channels at each step - a push notification first, then SMS and WhatsApp. Every incident gets handled, even if someone's phone is on silent or they slept through the notification. Acknowledge to stop escalation, or let it keep climbing to the catch-all so nothing ever ends in silence.
Every roster publishes a read-only calendar feed, and each person gets one covering their own shifts across every rota they are in. Subscribe from Google Calendar, Apple Calendar, or Outlook and on-call weeks sit alongside everything else that decides whether you book a weekend away. The feed follows overrides, so it shows who is actually covering rather than the raw rotation.
Off-duty should mean off-duty. Each person sets their own quiet hours and timezone, and during that window the personal channels aimed at them - mobile push and Telegram - are held back rather than delivered. Critical alerts can be allowed through, so it stays quiet for warnings and loud for outages. Suppressed pages are recorded rather than dropped, so "why did nobody get me" always has an answer.
Planned work produces the same alerts as an outage. Open a maintenance window over the whole account, specific integrations, or specific monitors, and alerts raised inside it are still recorded and still visible - they just stop notifying and stop escalating. The window has an end time, so unlike a paused monitor it closes itself, and the record of what happened during the work survives.
Not every alert deserves the pager. A round-robin escalation level hands one alert to one member of a roster and moves on to the next member for the alert after that, so a stream of low-severity work is shared out instead of landing on whoever happens to be on call. The pick is recorded per alert, so follow-ups and repeats reach the same person.
Pages reach Slack, Telegram, WhatsApp, SMS, email, and the mobile app. Acknowledging in any one of them stops the escalation everywhere, so nobody further up the chain gets woken for an incident that is already being handled.
Per-seat on-call tools charge you for every engineer you add to the rotation, which is exactly the wrong incentive - it pushes small teams to keep the rota short and burn the same two people out. AlertKick's bill tracks your infrastructure instead, and your whole team is included on every plan.
Install one agent. Get security monitoring, infrastructure alerts, and on-call management - all included with every host.