Aglet

Verify worker shutdown before and during active job steps

Verification must cover the worker phases that decide whether restart is safe. Stop at a safe point and during an active step, then restart the work. Confirm the durable checkpoint, claim owner, terminal state, and effects all tell the same recovery story.

Check whether the outcome improved

  1. Verify graceful drain

    Signal shutdown while the worker is idle or between jobs. Confirm it stops claiming new work and records a clean drain state. Use a queued job to verify another worker or later start can own it without creating a duplicate claim.

  2. Verify safe-point interruption

    Stop an active job just before its next effect and restart it. Confirm the checkpoint and requeue or review state identify the next safe action, with earlier effects retained once. Read durable claims instead of trusting worker logs.

  3. Verify mid-step recovery

    Interrupt the job during a known active step and observe its eventual state after restart. Confirm the step is completed, retried safely, or held for review according to policy. Reconcile output and side-effect identities to expose repeats or omissions.

What to carry forward

Accept when drain, safe-point, and mid-step cases have explicit ownership, checkpoint, terminal, and effect outcomes, with no lost or duplicate work. Reopen verification if a claim survives without an owner, restart skips a step, or a repeated effect lacks reconciliation.

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