A queue of batch jobs can need many machines briefly and very few once the work finishes. OpenAI's historical Kubernetes EC2 autoscaler adjusted AWS machine groups according to waiting jobs and their resource requirements. It was intended to supply worker capacity while accounting for constraints such as the kind of machine a job could use. The project is retired; understanding that capacity-ownership role is the starting point for replacing an existing installation.
Current guidance
The openai/kubernetes-ec2-autoscaler README states that the project is archived and no changes will be merged. It was designed for batch workloads running on EC2-backed Kubernetes capacity. The same source recommends Kubernetes Cluster Autoscaler, so the historical implementation should not be presented as a current supported autoscaling choice.
Inventory the AWS Auto Scaling Groups, pending-workload assumptions and scale-down behavior before transferring ownership. Match the replacement controller to the Kubernetes version and AWS node-group configuration. Batch jobs with node selectors, GPU requests or zonal volumes can remain unschedulable even when an autoscaler successfully adds a different kind of node.
Rehearse capacity limits, an unavailable instance type and interruption of a long-running job. Confirm checkpoints, retries and termination behavior so cost optimization does not silently discard work. Only one controller should own desired capacity for a given group, and its IAM permissions should be restricted accordingly. The original repository remains useful for understanding an existing deployment; its historical operating experience is not a current compatibility or reliability guarantee.
Historical upstream link check · 2026-10-09
The recorded upstream address responded successfully (HTTP 200) on 2026-10-09. GitHub marks openai/kubernetes-ec2-autoscaler as archived. This confirms the repository's read-only archive state; any successor or supported distribution needs separate evidence. Link availability does not certify the historical installation instructions or current security support.
Website availability is separate from project, chart and image support. Use the current guidance and primary sources on this page to assess the distribution.
Historical Kubedex content
Original publication: 2018-09-27T10:05:52+00:00. Preserved for context. Commands, versions, prices and results below reflect the original research.
kubernetes-ec2-autoscaler is a node-level autoscaler for Kubernetes on AWS EC2 that is designed for batch jobs. Kubernetes is a container orchestration framework that schedules Docker containers on a cluster, and kubernetes-ec2-autoscaler can scale AWS Auto Scaling Groups based on the pending job queue.
The key features are:
- Scaling on flexible resource requirements: the autoscaler determines the resources it needs from the pending job queue, and scales up the appropriate ASGs while respecting job constraints such as node selectors
- Multi-Region Support: the autoscaler can detect scaling errors and overflow to secondary AWS regions
- Draining nodes on scale in: the autoscaler makes sure to not kill in-flight jobs
Architecture
The post kubernetes-ec2-autoscaler appeared first on kubedex.com.
Sources & further reading
Spotted something that needs another look?
Help improve this page →