How to monitor a cron job (and know when it silently fails)

The problem with cron is that nothing happens when it fails. No page goes down, no error is shown; the backup just is not there when you need it.

The heartbeat pattern

Give each job a unique URL. The job calls it on success. A monitor expects the call every N hours; if it is late, you get an alert.

0 3 * * * /usr/local/bin/backup.sh && curl -fsS -m 10 https://hb.uptimepush.com/p/TOKEN

Add ExecStartPost=/usr/bin/curl -fsS https://hb.uptimepush.com/p/TOKEN to the service unit.

Append the curl call to the container command with && so it only runs on success.

Add a final step that curls the URL, with the token stored as a secret.

Set a grace period slightly longer than the job's longest normal run, so a slow night does not page you.