Comment by adityamishra241
6 hours ago
The pulse idea is neat. How do you handle jobs where the expected runtime is longer than the heartbeat interval?
6 hours ago
The pulse idea is neat. How do you handle jobs where the expected runtime is longer than the heartbeat interval?
Good question! There's no such thing as a "running" task/job at the moment (i.e., a start signal). The closest thing is lengthening the grace period, but that also means that the notification is delayed, so it's not really a solution. This is in the backlog.
The usual fix is a second ping at the start of the job: the schedule check watches the start ping, and a separate “max runtime” check watches the gap between start and finish. You get alerted promptly either way without inflating the grace period.