Aglet

Document the batch window and leftover-work contract

A window is a contract for what work may start, what happens to an item still running, and how the next occurrence treats leftovers. Write those rules with exact timestamps and ownership. Record service and dependency measures separately so future workload changes can be diagnosed without guessing from a late finish.

Keep the lesson for the next incident

  1. State cutoff semantics

    Define whether the boundary prevents new items, interrupts current work, or allows a documented drain. Name the terminal or partial status and the checkpoint that proves ownership. Include how due times are computed so an unrecorded timezone shift does not look like throughput loss.

  2. Keep boundary fixtures

    Retain normal, cutoff-crossing, and next-occurrence workloads with expected item coverage and outputs. Re-run them after changing batch size, dependency behavior, worker capacity, schedule interval, or checkpoint handling. Keep the schedule revision and worker pool recorded with each case so later timing changes remain attributable.

  3. Add recurrence review

    Watch for remaining items without a checkpoint, repeated cutoff crossings for one shape, and outputs updated after the next occurrence begins. Review the timeline before changing the window. Assign a contract owner when a new drain or overlap state is introduced.

What to carry forward

The durable lesson is an explicit cutoff, leftover ownership, and next-occurrence policy supported by boundary fixtures. Keep service-rate and dependency evidence distinct. Reopen review when workload shape, worker capacity, schedule computation, or checkpoint semantics changes the chance or meaning of an overrun.

Technical background: Celery documentation.

Keep the decision with the work.

Use a Work Item in Aglet to record the problem, the evidence you have, and the next decision. Add an owner and priority, then keep updates in the discussion so the next person can follow the reasoning.

Create an account See the product workflow