EMR cluster should use Spot instances for task nodes to reduce cost
What it checks
EMR cluster should use Spot instances for task nodes to reduce cost
Where it applies
Terraform: aws_emr_cluster
CloudFormation: AWS::EMR::Cluster
How to fix it
- Add Spot task capacity with the separate aws_emr_instance_group resource: cluster_id pointing at the cluster, bid_price set to the On-Demand price for interruption resistance. Task nodes hold no HDFS state and tolerate Spot reclaims.
- Where interruption tolerance allows, set bid_price on core_instance_group as well, so the core nodes also run on Spot. Either that or the aws_emr_instance_group above silences this finding.
Frameworks
- AWS Well-Architected COST.7
These mappings are advisory, not an attestation of compliance.