When you remove a server from rotation (for deployment or failure), what happens to in-flight requests?
Without draining: Requests get dropped. Users see errors.
With connection draining: Load balancer stops sending NEW requests but allows existing connections to complete.
Typical drain timeout: - seconds. Long enough for most requests to finish. Short enough to not delay deployments.
In Kubernetes: This is handled by the readiness probe and termination grace period.
Mention connection draining when discussing deployments or failover. It shows you think about production operations.