Operational Excellence Operational Excellence Operational Excellence

Chọn baseline cho traffic, cache hit/miss, 4xx/5xx và latency trước khi đặt alert. Establish baselines for traffic, cache hit/miss, 4xx/5xx, and latency before creating alerts. Establish baselines for traffic, cache hit/miss, 4xx/5xx, and latency before creating alerts.

Không cần bài trước. No prerequisites. No prerequisites.

Dùng structured logs, request ID và traces để nối lỗi user-facing với code path. Use structured logs, request IDs, and traces to connect user-facing errors to code paths. Use structured logs, request IDs, and traces to connect user-facing errors to code paths.

Triage WAF, DDoS và bot events; export log khi cần retention hoặc correlation bên ngoài. Triage WAF, DDoS, and bot events; export logs when you need retention or external correlation. Triage WAF, DDoS, and bot events; export logs when you need retention or external correlation.

Theo dõi model, latency, error và cost signal mà không log prompt nhạy cảm không cần thiết. Track model, latency, error, and cost signals without unnecessarily logging sensitive prompts. Track model, latency, error, and cost signals without unnecessarily logging sensitive prompts.

Khi error spike, kiểm tra Cloudflare Status cùng metric và deployment timeline trước khi kết luận root cause. When errors spike, check Cloudflare Status alongside metrics and deployment timelines before concluding root cause. When errors spike, check Cloudflare Status alongside metrics and deployment timelines before concluding root cause.

Phân biệt attack, false positive và app regression; preserve evidence trước khi thay policy. Separate attacks, false positives, and app regressions; preserve evidence before changing policy. Separate attacks, false positives, and app regressions; preserve evidence before changing policy.

Define owner, severity, update cadence, rollback criteria và post-incident actions trước incident tiếp theo. Define owners, severity, update cadence, rollback criteria, and post-incident actions before the next incident. Define owners, severity, update cadence, rollback criteria, and post-incident actions before the next incident.

Test preview URL, define production checks và giữ rollback path rõ ràng trước release. Test preview URLs, define production checks, and keep a clear rollback path before release. Test preview URLs, define production checks, and keep a clear rollback path before release.

Tách config theo environment; secret không nằm trong source; log ai thay đổi gì và khi nào. Separate configuration by environment; never place secrets in source; record who changed what and when. Separate configuration by environment; never place secrets in source; record who changed what and when.

Version asset, purge đúng scope và kiểm tra cache behavior sau frontend/API release. Version assets, purge the right scope, and verify cache behavior after frontend/API releases. Version assets, purge the right scope, and verify cache behavior after frontend/API releases.

Định nghĩa health signal có ý nghĩa cho user và test failover trước khi cần dùng. Define health signals meaningful to users and test failover before you need it. Define health signals meaningful to users and test failover before you need it.

Kết hợp proxy, DDoS protection, WAF/rate limit và origin hardening để giữ service available. Combine proxying, DDoS protection, WAF/rate limiting, and origin hardening to keep services available. Combine proxying, DDoS protection, WAF/rate limiting, and origin hardening to keep services available.

Mở rộng policy hoặc Zero Trust deployment theo nhóm nhỏ, feedback loop và rollback plan. Expand policy or Zero Trust deployments through small groups, feedback loops, and rollback plans. Expand policy or Zero Trust deployments through small groups, feedback loops, and rollback plans.

Đo LCP/FCP/CLS, hit ratio và origin request để ưu tiên công việc performance. Measure LCP/FCP/CLS, hit ratio, and origin requests to prioritize performance work. Measure LCP/FCP/CLS, hit ratio, and origin requests to prioritize performance work.

Review usage, product limits và cost driver trước khi scale workload hoặc bật add-on. Review usage, product limits, and cost drivers before scaling workloads or enabling add-ons. Review usage, product limits, and cost drivers before scaling workloads or enabling add-ons.

Lập monthly review cho changelog, incidents, policy exceptions, SLO và backlog reliability. Create a monthly review for changelog, incidents, policy exceptions, SLOs, and reliability backlog. Create a monthly review for changelog, incidents, policy exceptions, SLOs, and reliability backlog.