hard20 pts · first answer countscost-optimizationkubernetesreliability

Running workloads on spot

You want to cut compute cost by moving Kubernetes workloads to spot instances. Which approach is sound?

A.Move every workload, including the single-replica database, to one cheap spot instance type to save the most.B.Run only the Kubernetes control plane on spot, because it is managed and recovers automatically when reclaimed.C.Put stateless replicated apps on varied spot types and zones with graceful drains; keep critical singletons on-demand.D.Disable all PodDisruptionBudgets so that spot nodes can be drained and removed faster when reclaimed.

← Huge NAT gateway charges First move in a post-deploy outage →