Decide who gets paged first, who's next if nobody acks in ten minutes, and which channel catches the 3 a.m. call when the first person sleeps through it. Hoppla runs its own on-call engine: timezone-aware rotations, ordered steps, and up to ten repeats while nobody acks. Once a human acknowledges, no further step fires. On-call schedules are included from Starter, with no per-seat PagerDuty bill.
incident +0m on-call(eu) · +10m backup · +20m repeat · ACK→halt
Fictional values. Not a live measurement.
An escalation policy is an ordered list of up to ten steps, each with a wait time; the waits add up from the moment the incident opened. Step one fires. If nobody acknowledges, the next step fires once its wait elapses, notifying either a channel directly or whoever's on call right now. On-call schedules hold the rotations: a user and a channel, with optional day-of-week and time windows, all evaluated in the schedule's own timezone. Once someone acknowledges or resolves the incident, no further steps fire.
When an incident opens and a matching escalation policy exists, Hoppla records the start time and starts tracking. Escalation only ever runs for incidents that are still unacknowledged and unresolved. Nothing fires for an alert a human already grabbed.
Every minute, the scheduler checks elapsed time against each step's cumulative wait_minutes from incident start. Step one might page at 0 minutes, step two at 10, step three at 25. As a threshold passes, that step notifies its target: a channel directly, or the people the on-call schedule says are on duty.
A step pointed at a schedule resolves the current rotation in that schedule's timezone. It matches the day of week and the HH:MM window, handles windows that cross midnight, and pages everyone on duty on their own channel or the step's. If nobody is on duty, it falls back to the step's channel, as long as the step has one.
Acknowledging or resolving the incident marks escalation done, and no further steps fire. If the whole policy runs out with nobody responding, repeat_count re-runs it from the top, up to ten times, so an incident no one answers comes back instead of going quiet.
One missed notification is how an outage runs for hours unnoticed. When a policy burns through every step with no ack, repeat_count re-runs it from the top, up to ten times, and the incident keeps paging until a human grabs it. Silence doesn't read as handled.
Rotations resolve in the schedule's own timezone, match day-of-week and HH:MM windows, and handle shifts that cross midnight. Whoever's actually on duty gets paged, not whoever happened to sit first in a flat list.
PagerDuty and Opsgenie bill per responder. Hoppla's on-call is part of the plan, from Starter. Starter includes five team members and Pro has no member limit, so there's no per-responder fee gating who you can page.
You are the entire rotation, so the failure mode isn't 'who's on call', it's sleeping through the first alert. Build a two-step policy: step one pings your phone, and if you haven't acked in ten minutes, step two hits a louder backup channel. Set repeat_count so a real incident comes back instead of dying in a notification you swiped away half-asleep. Ack from bed and no further step fires.
Two of you share the after-hours phone, and you'd rather not sign a PagerDuty contract to coordinate it. One schedule, one rotation per person split by weekday, plus a fallback step to a shared channel if neither acks in fifteen minutes. The person on duty takes the incident, the other isn't woken unless the fallback fires, and there's no per-seat bill for something this small.
Responders sit in different timezones, and clients expect coverage that doesn't lapse at the seams of a shift. Each schedule is evaluated in its own timezone with real day-of-week and time windows, windows across midnight included. Stack a multi-step policy on top (primary on-call, then a team lead at 20 minutes, then a manager channel) with repeats while nobody acks, so a 4 a.m. incident keeps paging instead of sitting unowned while everyone assumes someone else has it.
Managed on-call can be a premium tier you sell, without reselling PagerDuty and eating its per-seat cost. Hoppla's escalation engine is white-labeled and built in, so you hand each customer schedules and policies under your own brand and domain. The pages, the acks, all of it wears your name. You sell 24/7 incident response as your product, with no seat-based vendor invoice eating the margin.
Import the monitors you already run in another tool. What comes across depends on the provider. Run both side by side for 24 hours or 7 days before you switch.
The importer opens in your workspace once you have access.
Get early accessNo. This is Hoppla's own on-call engine. The schedules, rotations, ordered steps, repeat logic and acknowledge-to-halt all run inside Hoppla, fed by the same monitors raising your alerts. You can still point a step at a PagerDuty service as one of your channels if you want it in the loop, but you don't need a PagerDuty or Opsgenie subscription to get full escalation.
Acknowledging or resolving the incident marks escalation done, and no further steps fire. Escalation only ever runs while an incident is both unacknowledged and unresolved, so once a responder grabs it, the chain stops. Acknowledging is all it takes.
That's what repeat_count is for. When a policy runs through every step with no ack, it re-runs from the top, up to ten times, at a repeat interval you set. The incident keeps paging on the schedule you defined instead of falling silent, so an alert everyone ignored is less likely to turn quietly into a multi-hour outage.
Yes. Each schedule carries its own timezone, and rotations take an optional day-of-week list plus an HH:MM start/end window evaluated in that timezone. Windows that cross midnight, say 22:00 to 06:00, are handled. One detail: the day-of-week list is checked against the local day at that moment, so the hours after midnight count toward the next day.
Checks, alerts and status pages in one workspace. Now in early access: join the list for an invite.