Using spot nodes in your Kubernetes cluster can be intimidating due to their lack of availability guarantees. Kubecost’s Spot-Readiness Checklist is here to give you more confidence. The checklist investigates your public cloud Kubernetes workloads to identify candidates for safe scheduling on spot instance types, which can save you up to 90% on cloud resource costs. Kubecost automatically performs a series of checks on your AWS (EKS), Azure (AKS), and Google Cloud (GKE) clusters using your workload configurations to determine readiness. It then estimates the savings impact from making the transition to Spot.
What Are Spot Instances and Why Use Them?
Spot instances are spare compute instances that public cloud providers offer to customers at a deeply discounted rate—potentially up to 90% cheaper. However, spot nodes vary in their availability and pricing, depending on the supply and demand of compute resources at a given time and fluctuate per instance size, instance family, and deployment location. Once the demand for a particular instance type increases, spot instances may receive an interruption notice and spin down within a short shutdown window (usually a few minutes). For this reason, spot resources are best used for fault-tolerant and flexible applications like Spark/Hadoop nodes, microservices that can be replicated, etc.