Autom Mate Team · Practical example
A create request times out. Did the destination reject it, or did it create the record and lose the response? Retrying immediately can create a duplicate.
Design the retry around the business operation, not just the error message.
- Give the operation a stable reference. Use the source record ID and intended operation. Keep the same reference across retries.
- Store the destination result. When a create succeeds, retain its returned ID and link it to the source.
- Check after an uncertain failure. Query the destination for that reference before another create. Use its idempotency mechanism where the API supports one.
- Retry only the appropriate failures. Use bounded retries for temporary failures. Missing required fields need correction, not repeated submission.
- Verify the final record. Confirm the destination ID and expected fields. If the outcome remains uncertain, route it for review.
Example: a customer-record create times out, but a lookup finds the new record. Store its ID and continue instead of creating again.
Done when: a repeated operation leaves one correct destination record.
Autom Mate’s error-handling guide explains retry behaviour; duplicate protection still belongs in your flow design.
Which system gives you the most ambiguous timeout responses?