SUSE Rancher can manage AWS EC2 Spot Instance based Managed Node Groups with EC2 Auto Scaling

SUSE Rancher & Amazon Web Services Spot Instance-based managed node groups for incredible savings and familiar management – a not-so-new game-changer from SUSE.

SUSE Rancher can manage AWS EC2 Spot Instance-based managed node groups with EC2 Autoscaling

AWS Spot-priced Instances have been around a while and boast a whopping 90% savings from comparable On-Demand instance consumption. The rub – while the instance is identical in every way to a standard On-Demand instance, or Reserved Instance, it may be terminated at any time by AWS as EC2 capacity requires. Spot is therefore best suited for distributed computing with ephemeral, scalable workloads.

Enter managed node groups. What better example to leverage Spot instance savings than with ephemeral workloads designed to withstand vanishing servers? While Kubernetes scales your application, managed node groups scale the underlying infrastructure as a service (IaaS.) EKS Managed Node Groups leverage AWS Autoscaling configuration to ensure that your required service level is maintained, despite Spot Instances disappearing on you.

This solution can be deployed directly from SUSE Rancher. You can leverage your familiar single pane of glass to save 90% from your workload cluster infrastructure by deploying your workload clusters using EC2 Spot Instances by simply selecting the “Spot” check box in the EC2 Cluster deployment workflow within Rancher.

SUSE advises the following configuration scenario: SUSE Rancher can first be deployed on its own management cluster or directly into Docker on SUSE Linux, into EKS, etc. At this point, leverage SUSE Rancher to deploy a small on-demand or reserved instance managed node group of 1 or 3 nodes to support the control plane with at least one worker node using standard instance types, rather than Spot. Additional reserved or on-demand instances should be deployed for mission-critical workloads that cannot tolerate a reduction in capacity. Determine your application’s low water mark and deploy this minimum node group accordingly. Finally, deploy one of more managed node groups using the Spot feature. Repeat this procedure by leveraging varying instance types and at varying bid pricing to minimize the risk surface to your application.

This final node group can account for most of your cluster, and here is why – We leverage different instance types with the same basic capacities. Example: If the requirement states 8GB Memory, 2vCPU, etc., then this can be accomplished using 2 or more different instance types launched with the Spot option, further decreasing the chances of losing too much of your cluster to an EC2 capacity event affecting a particular type. Larger sizes can be used, providing even more diversity because of course, you are saving up to 90%. The Autoscaling Group created from the Rancher User Interface ensures that EC2 will manage the cluster size.

Reference the SUSE Rancher official documentation or reach out to SUSE any time at aws@suse.com for more information on how to manage your EC2-based Kubernetes clusters.

References

EC2 Spot Instance best practices

Best Practices for SUSE Rancher Server

Auto Scaling with multiple instance types