Rust Idempotent Workflows

Rust Idempotent Workflows

Use this skill to make Rust workflows reliable when clients retry, workers crash, requests race, and external side effects cannot be rolled back.

Core Workflow

1. Write the failure story before writing code. List what happens if the process crashes after each database write and after each external side effect.
2. Identify the idempotency boundary: HTTP endpoint, command handler, queue job, webhook, or external callback.
3. Choose the idempotency strategy:
   • Natural idempotency from a unique business key.
   • Client-provided idempotency key with save-and-replay.
   • Server-generated deduplication key.
   • Outbox or queue table for side effects.
4. Put state transitions and idempotency records in the same transaction when they must agree.
5. Make duplicate concurrent requests deterministic. Use unique constraints, locks, serializable transactions, or explicit conflict handling.
6. Treat external side effects as at-least-once unless the provider guarantees more. Store enough state to retry or reconcile.
7. Add tests for duplicate requests, retry after timeout, concurrent requests, crash windows when practical, and worker retry behavior.

Idempotency Key Rules

Read references/idempotency.md before adding idempotency keys or save-and- replay behavior.

• Bind the key to the authenticated user or tenant.
• Store a hash of the request payload when replaying responses.
• Return the original response for exact replays.
• Reject reuse of the same key with a different payload.
• Expire records only after product and provider retry windows have passed.

Background Work Rules

Read references/background-workers.md when adding queue tables, polling workers, retries, or external side effects.