architect-error-handling

Design error handling for distributed transactions by defining catch locations across controllers and services, applying state guarantees like idempotency and compensating actions, and setting retry policies with backoff for partial-failure recovery.

Handle partial failures in a multi-step service flow by implementing state recovery strategies such as sagas, compensating actions, and idempotency keys to ensure failures either roll back safely or allow systematic recovery.

Decide where to catch errors across service layers by establishing explicit error boundary decisions that determine whether to catch, translate, or let errors propagate from controllers through services to infrastructure.

Categorize API errors for callers by defining caller-facing error formats and retry rules that explicitly distinguish between retryable and non-retryable failures, ensuring consistent client behavior and observability.

You need compensating actions and idempotency keys when designing error recovery for multi-step operations that require safe rollback or systematic recovery, ensuring operations remain atomic and preventing duplicate side effects during retries.

Document error handling decisions for a pull request by producing a short PR attestation that summarizes catch locations, state guarantees, retry policies, and caller-facing error formats to communicate design choices clearly to reviewers.