Shrutam CCA-F Crash › Domain 5 › d5-l07-reliability-patterns Hinglish English →

Reliability Patterns — Retries, Fallbacks, Circuit Breakers

Domain 5 · 15% ~12 min Hinglish narration

Audio-only (commute / mobile data)

Same Saavi narration, smaller file. Opus 48k preferred — auto-selected by your browser.

Scenario anchor

Aap Aaranya IT BFSI ke ek senior architect hain — client hai ASX-India, aur unka naya trade-confirmation pipeline Claude API se long-context contract analysis kar raha hai. Market open hone ke pehle 9:14 AM pe, Claude API ne throttling shuru kar di — HTTP 429s aa rahe hain. Aapka na-to retry logic theek hai, na koi fallback strategy. Pipeline halt, ASX-India operations team escalation pe, aur aap incident call pe hain. Is lesson mein hum exactly yahi fix karenge — production-grade reliability patterns jo Claude ke saath kaam karte hain.

Key Takeaways

On Claude API 429 or 529 errors, always use exponential backoff with jitter — not fixed-interval — to prevent thundering-herd amplification across concurrent agent instances.
Implement a three-state circuit breaker (CLOSED → OPEN → HALF-OPEN) so the system can self-heal after transient Claude API rate-limit windows without manual ops intervention.
Design fallback tiers explicitly: first a cached prior response, then a reduced-context re-invocation, then a graceful degradation message — never a naked failure propagated upstream to the ASX-India-tier orchestrator.
Memory anchor — treat Claude API reliability exactly like a service-mesh circuit breaker: state transitions, probe requests, and fallback routing are all first-class design concerns, not afterthoughts.

← Previous lesson Same lesson in English Take Mock Exam 1 → Back to overview