Track 03

Software & production

Design work you can retry, inspect, and recover after a failure.

Start here

Design for an interrupted run

  1. Give the work an owner. Define what a worker claims and when another worker can take over.

  2. Make retries safe. Record enough to repeat a step without writing the same change twice.

  3. Report what completed. Keep the last confirmed checkpoint and name any missing or stale results.

Explore the lease queue lab

Worked examples

References

In development

Explore all five tracks