Kubernetes gives us powerful tools to build resilient systems, but simply running multiple replicas doesn't automatically make an application highly available. What happens when a node is upgraded, a pod is terminated, or a new version is rolled out while users are actively making requests? We'll explore what it really takes to keep an application highly available when Kubernetes performs maintenance, workloads are disrupted, or failures occur. We'll look at the responsibilities on both sides: what Kubernetes can do to keep workloads available, and what the application itself needs to do to handle disruption gracefully. The goal is to understand how these two layers work together to build applications that remain available when the platform changes or things inevitably fail.