Alibaba: How Do You Implement Service Discovery and Governance at Scale?

I'd build the governance layer on Alibaba's proven stack. For service discovery, Nacos serves as both registry and config center — services register on startup with health-check endpoints, and consumers get real-time push notifications of instance changes. Load balancing uses a weighted strategy that factors in each instance's response time and error rate, not just round-robin. For traffic governance, Sentinel enforces rate limits at the API gateway and within each service. Circuit breakers monitor error rates over sliding windows — when failures exceed the threshold, the circuit opens and requests are fast-failed or routed to fallback logic. System adaptive protection kicks in when CPU or load exceeds safe thresholds, automatically shedding low-priority traffic. To prevent cascading failures, I'd implement bulkhead isolation: critical and non-critical services run in separate thread pools, so a slow recommendation service can't exhaust threads needed for checkout. Distributed tracing via OpenTelemetry gives end-to-end visibility, and automated alerting triggers when P99 latency or error rate anomalies are detected.