Missed, vanished, and never-ran jobs
A failing job that reports its error is easy to see. The harder cases are jobs that stop reporting: a run that never happens, a run that starts and hangs, a new job that never runs at all, and a job that disappears from your scheduler. This page explains how CronRadar finds each one.
Missed runs
Every report tells CronRadar when the next run is due: the next time the job's cron schedule fires after the report arrived, in the application's time zone (UTC unless set). CronRadar checks for overdue jobs every minute.
- Late: once the due time passes, the job turns
warning. No alert is sent yet. - Missed: when the grace period ends with no report, and no sooner than two minutes after the due time, the job turns
criticalwith the resulttimeout, and CronRadar sendsInitialFailure. - Still missing: while the incident stays open, CronRadar sends a
ContinuedFailurereminder every 24 hours. Acknowledging the incident in the console stops the reminders. - Back: the next completion or heartbeat closes the incident and sends
Recoveredwith the downtime. A start event alone does not close it.
The grace period absorbs normal lateness. When you create a job in the console without one, CronRadar uses half the schedule's interval, at least 90 seconds; console values range from 60 seconds to 24 hours. Jobs created by their first report or through framework sync default to 90 seconds unless the integration sends another value. See Job statuses and timing and Cron schedule syntax.
Stuck runs
When a job reports a start, CronRadar marks it as running. If no completion or failure arrives before the next run is due plus the grace period, CronRadar treats the run as stuck: the job turns critical and CronRadar sends InitialFailure, explaining that the run started but did not finish in time. There is no separate maximum-runtime setting; the schedule sets the window.
Stuck-run detection needs start events. A job that only sends heartbeats can be detected as missed, but not as stuck. Framework integrations and SDK wrappers send start, completion, and failure for you.
Jobs that never ran
A job created in the console or discovered through framework sync starts as pending. CronRadar knows when its first run is due. Every ten minutes it checks for pending jobs whose first run plus grace has passed with no report, and sends NeverExecuted. The alert repeats every 24 hours until the job reports or you pause it.
NeverExecuted does not open an incident, so the job's first report does not send Recovered. A job created by its own first report never needs this check.
Jobs that vanish from your scheduler
The Hangfire integration sends CronRadar the full list of recurring jobs in Hangfire's registry at startup and every 15 minutes by default. Each list is a reconciliation: CronRadar compares it with the jobs that came from the same source. Other integrations behave this way only where their setup guide says so.
When a job that was in an earlier list is missing from a complete, valid list:
- CronRadar sends one
Vanishedalert through your alert channels, then pauses the job. Its health showspausedwith the reason "No longer declared by the connected scheduler." - When several jobs leave in the same list, for example after a deploy that broke the integration, CronRadar sends one combined
Vanishedalert that names one job and counts the rest, not one alert per job. - The alert is sent once per disappearance. There are no daily reminders, and jobs that were already paused send nothing.
- Its history is kept; nothing is deleted.
- Any open incident for it is closed.
- It stops counting toward your monitor total.
If the job comes back in a later list, CronRadar resumes it, provided your plan or budget has room; over the budget it is held until the budget allows. No Recovered alert is sent: a job back in the registry has not run yet, and from then on its runs are watched as usual. A job you paused yourself is never resumed by sync.
CronRadar skips removal entirely when a list contains an invalid entry, so one bad declaration cannot pause the rest, and it refuses a reconcile request that has no monitors list with 400. Jobs created in the console or by their first HTTP report are never paused by sync.
To stop monitoring a job on purpose without removing it from the scheduler, use the integration's opt-out, such as [SkipMonitor] on Hangfire.
Jobs created by their first report
A report for an unknown job key that includes ?schedule= creates the job, so a deleted job registers again on its next run. Without a schedule, the report returns 404. When the job would push your bill over your monthly budget, the report is still recorded and the job is created held paused until the budget allows. At the Free limit, the report returns 402 and no job is created; reports for existing jobs are still recorded.