• Asp.Net Core
• 2
• September-20-2025
• by Ihsan Ullah

Q: How to keep auth stable?
A: Choose OAuth2 with refresh tokens. Store secrets in a vault. Request least-privilege scopes. Build a small token service that refreshes before expiry and exposes tokens to callers. Log vendor error and error_description. Rotate keys on a schedule.

Q: How to survive rate limits?
A: Read vendor docs for exact quotas. Respect Retry-After. Add client-side throttling per API key. Batch or cache reads. Spread traffic across minutes rather than bursts. Create dashboards for 4xx/429 rates.

Q: How to handle timeouts/flaky networks?
A: Set short connect/read timeouts. Retry only idempotent operations. Use a circuit breaker so failing endpoints open quickly and recover gradually. Provide graceful fallbacks and user-visible status.

Q: How to avoid duplicate effects on retries?
A: For each “logical operation,” generate a unique operation ID. Send it with the request and store outcome against that ID. On retry, return the stored outcome.

Q: How to avoid breaking changes?
A: Pin the API version in every request. Subscribe to vendor change logs. Run nightly contract tests that hit the sandbox and verify required fields. Track deprecation headers and schedule upgrades.

Q: How to page reliably?
A: Prefer cursor/tokens over offsets. Loop until no next token. De-duplicate by stable IDs. Persist the last token so jobs can resume after failure.

Q: How to secure webhooks?
A: Put webhooks behind an allowlist and a separate ingress endpoint. Verify vendor signature and timestamp. Enforce idempotency using the event ID. Accept, queue, then process to keep the endpoint fast.

Q: How to handle currency/time/locale?
A: Always send explicit currency codes and ISO-8601 timestamps with timezone. Convert to UTC internally. Render for users in their locale on output.

Q: How to cut payload and latency?
A: Use vendor fields-filtering and bulk endpoints. Compress over the wire. Parallelize reads with sane concurrency caps. Cache stable GETs with TTLs and cache keys that include all query params.

Q: How to prevent secret leaks?
A: Centralize logging with redaction. Never log headers or bodies from auth paths. Use short-lived credentials. Rotate and monitor access to the vault.

Q: How to meet legal/compliance needs?
A: Sign the DPA. Verify data residency and subprocessors. Classify data you store. Minimize retention. Provide deletion paths and audit trails.

Q: How to test without breaking prod?
A: Use the vendor sandbox. Seed deterministic fixtures. Record/replay happy and failure paths. Add contract tests that validate required fields and enums. Run chaos tests for timeouts and 5xx.

Implementation flow (fits ASP.NET MVC5 / EF6 / Hangfire)

1. Discovery

• Read API docs: auth, quotas, pagination, webhooks, versioning.

• List endpoints, payload shapes, and SLAs.

• Decide sync vs async: direct calls in requests vs background jobs.

1. Design

• One client per vendor with its own base URL, headers, and policies.

• Token service owned by you.

• Idempotency model: operation IDs + result store.

• Caching strategy for GETs.

• Webhook ingress: verify, enqueue, process.

1. Observability

• Metrics: latency, success %, error by type, retries, breaker state, rate-limit hits.

• Tracing: correlation IDs from user → API call → webhook.

• Logs: structured with redaction.

1. Data model

• Tables for: outbound requests, idempotency keys, pagination cursors, webhooks received, vendor objects cache, error catalog.

• Use EF6 migrations to add these.

1. Job orchestration

• Hangfire recurring jobs per endpoint for sync tasks.

• Queues per vendor priority.

• Dead-letter queue for poison events.

1. Error strategy

• Map vendor errors to actions: retry, fix input, or alert.

• Backoff with jitter.

• Kill switch per endpoint if error rates spike.

1. Security

• Secrets in a vault. Least privilege. IP allowlists where offered.

• Webhook signature verification.

• Access reviews and key rotation calendar.

1. Compliance

• Document data flows.

• Retention policy per data class.

• Incident runbook and contact list for the vendor.

1. Rollout

• Feature flag around the integration.

• Canary with 1–5% traffic.

• Budget caps for costed APIs.

• On-call playbook with clear rollback.

1. Maintenance

• Watch vendor status pages and change logs.

• Renew credentials before expiry.

• Quarterly disaster drills: token expiry, 429 storms, webhook outage.

Acceptance checklist

• Auth refresh works without user impact.

• Idempotent writes verified by replay test.

• Pagination completes and resumes after crash.

• Webhooks validated, queued, deduped, and ordered.

• Rate-limit compliance proven under load.

• PII redaction present in logs.

• Dashboards and alerts in place.

• Runbook links and on-call rotations set.