Run history, duration and stuck runs
CronRadar keeps a history of every reported run of a job: when it started, when it finished, how long it took, whether it succeeded, and whether it was late. A run that reports a start but no finish by the job's next due time plus its grace period is a stuck run, and CronRadar alerts on it. There is no separate maximum-runtime alert: the schedule sets the window.
How long history is kept
| Plan | Run history | Alert history |
|---|---|---|
| Free | 30 days | 90 days |
| Pay as you grow | 90 days | 180 days |
A daily cleanup removes older records using the billing owner's current plan. Moving to a plan with longer history can't restore records already removed; moving to a shorter one applies at the next cleanup. See Billing.
What each run shows
Each job's page in the console lists its recent runs. For every run it shows:
- Started and Finished, from the job's
startandcompleteorfailreports - Duration, the time from start to finish
- Expected, when the run was due, and Lateness, how far after that it started
- Attempts, how many tries the run took, from the retry number sent with each start
- Source, the integration that reported it, and Execution ID, which ties a run's reports together
Duration needs a start and a finish. Framework integrations and SDK wrappers send both. With the HTTP API, send start before the job and complete or fail after it; a heartbeat records a successful run with no duration. Send the same run_id with both reports so runs that overlap, or retries of one run, are paired correctly. Without a run_id, a start is paired with the next finish and the run is labelled as legacy evidence.
Stuck runs
When a job reports a start, CronRadar marks it as running. If no completion or failure arrives before the job's next due time plus its grace period, the run is stuck:
- The job turns
critical. - CronRadar sends
InitialFailure, saying the run started but did not finish in time. - While the incident stays open, a
ContinuedFailurereminder follows every 24 hours. - The next completion closes the incident and sends
Recovered.
The window comes from the schedule, so a job that runs every 15 minutes is stuck sooner than one that runs daily. There is no separate maximum-runtime or duration-threshold alert. A job that sends only heartbeats can be caught as missed, but not as stuck.
Missed, vanished and never-ran jobs covers the other ways a job stops reporting, and Job statuses and timing defines each status.