Posted On: May 24, 2021

You can now use Amazon EC2 On-Demand Capacity Reservations (ODCRs) while creating clusters in Amazon EMR. When rotating a long running cluster, you might have strict requirements on the instance types and Availability Zones for the new instances you provision. Using Capacity Reservations, you can get capacity assurance to complete the cluster rotation without interruptions. You can also use Capacity Reservations to provision a group of successive short-lived clusters for individual workloads such that when you terminate a cluster, the next cluster can use the Capacity Reservations.

By default, EMR provisions On-Demand instances using a lowest-price strategy, i.e., it uses the lowest-priced instances first to meet the capacity requirements. If the instances provisioned match the attributes of available open Capacity Reservations (instance type, platform, tenancy and Availability Zone), the open Capacity Reservations get applied automatically. Now, by using Instance Fleets with allocation strategy, you can provision EMR clusters which use open or targeted Capacity Reservations first and then, if required, fall-back to the lowest-price strategy. You can also choose to override the default use of open Capacity Reservations, even if the provisioned instances match their attributes.

You can cancel Capacity Reservations whenever you like. This will not affect your existing EMR clusters which are using the Capacity Reservations. They will continue to run the On-Demand instances until you choose to stop them. However, new clusters or additional instances provisioned in existing clusters will not be able to use the cancelled Capacity Reservations. Instead, they can use other available Capacity Reservations or the On-Demand pools. To get started, read Use Capacity Reservations with Instance Fleets and On-Demand Capacity Reservations, and watch the targeted Capacity Reservation demo.