Define successful completion before adding a heartbeat
UptimeRobot’s cron documentation describes a monitor that expects requests within an interval and grace period. Its setup guide explains sending requests from scheduled tasks. Unlike an external page check, the task initiates the signal.
Define the outcome first. For a catalog import, success might require a saved run record and a reviewed count of accepted records. For an export, it might require an output file that can be opened. Place the signal after that outcome, rather than at the start of the task or in a separate scheduler that runs regardless of the task result.
A job can finish normally while doing nothing useful. Keep output validation separate from the heartbeat and record the distinction in the runbook. For backups, schedule restore testing independently: an arriving ping does not establish recoverability.
Choose timing from the actual schedule
List the schedule, time zone, normal duration and the latest acceptable completion time. Then choose a grace period that allows normal variation while still exposing a missed delivery soon enough to act. The provider documents interval and grace controls; verify their current availability in the account you will use.
Illustrative example: a morning export due before an operations meeting needs an alert while someone can still recover it. A warning arriving after the meeting is less useful, even if it accurately detects a missed run. Daylight-saving changes, maintenance and unusually long inputs should appear in your timing notes. Avoid copying the shortest interval simply because it is available.
Keep this checklist
- Record the schedule and time zone.
- Identify the completion deadline and owner.
- Document planned maintenance handling.
- Check what happens if the signal cannot be delivered.
Test a missed run and rehearse recovery
In an approved test environment, omit a heartbeat, confirm the expected notification and restore the signal. Separately rehearse a task that fails before completion and one that completes its work but cannot send the check-in. These scenarios help the responder distinguish a failed task from a failed notification path.
Before rerunning a task, inspect its output and last successful run. A partially completed import may already have changed records. Use stable record identifiers and documented retry behavior to avoid duplicate output; do not treat an alert as authorization to replay every external action.
The response note should identify the task, schedule, last expected outcome, output location and escalation route. Keep monitor URLs in private configuration rather than public examples. Retain ordinary HTTPS certificate verification when implementing requests. We are proposing these controls, not reporting a tested integration.
Use a heartbeat for recurring work, not every page
A website can remain available while an overnight import stops. A recurring task can also complete while its public output fails to load. Use both task-completion and page checks when both outcomes matter, keeping their response notes distinct.
Start with one consequential job and review alert usefulness after a rehearsal. See the [UptimeRobot tool profile](https://softwarevista.online/software/uptimerobot) before evaluating the service, and our [automation recovery checklist](https://softwarevista.online/blog/automation-failure-monitoring-checklist) for broader retry planning. For missing public content, use the separate [keyword monitoring guide](https://softwarevista.online/blog/uptimerobot-keyword-monitoring). Check current plan requirements directly; this guide does not quote a price or promise uptime.
Common questions
Should a heartbeat be sent before or after a job?
For a completion check, send it after the validated successful outcome. A signal sent before the work cannot establish that the work finished.
Can a heartbeat prove a backup is usable?
No. Validate the backup output and perform a separate restore test. The monitor checks receipt of the signal.
Sources & verification
Source review: 2026-10-10. AI-assisted desk research of official documentation. SoftwareVista has not performed hands-on tests. The examples and checklists are proposed operating practices, not measured product results.
