Skip to slide
216 / 221

CH 18 · Reference Architecture and a Design Checklist · 11/16

Read as one page

Reliability

  • Is all model output parsed defensively, turning mistakes into recoverable signals? (Ch 13)
  • Are transient failures retried with capped backoff + jitter, terminal ones not? (Ch 13)
  • Are side-effecting operations idempotent? (Ch 13)
  • Is progress persisted for resumability, and each dependency classified fatal vs degradable? (Ch 13)
  • Do streams fail cleanly (error event + sentinel), with resource cleanup on disconnect? (Ch 8, 13)
  • Are there timeouts on every external call? (Ch 13)

Finished this idea?

Your learning progress stays in this browser.