3 / 4 Schedulers, Cron & Job Systems

Three Hours of Silence

A job runs every 5 minutes to recompute a rolling usage aggregate, and each run fully supersedes the previous one. The scheduler was down for 3 hours; on restart it enqueued all 36 missed occurrences at once and the database fell over. On-call wants the aggregate fresh as soon as possible. Which misfire policy fits?

A

Raise the scheduler's dispatch concurrency so all 36 runs execute in parallel and the backlog clears in a single 5-minute window

B

Coalesce the missed occurrences into one run that fires immediately, then resume the normal 5-minute cadence from the current time

C

Backfill all 36 occurrences but spread them with jitter across the following hour so the database sees a smooth load instead of a spike

D

Discard every missed occurrence and simply wait for the next scheduled slot, keeping the schedule aligned to its original grid